Twu31/bci-report

BCI Report

Every EEG decoding score, reported with the protocol that produced it.

Website · 中文 · Code · Data use & privacy · Every dataset · Every method · Data API

BCI Report Explorer: browse these results, or query them through MCP tools, in a Space that reads one pinned release of this mirror and names it in every answer.

121 reviewed measurements from public EEG datasets, each carrying the cohort, electrode count, evaluation mode, chance level and training budget that produced it. Core release research-preview-20260920, reviewed 2026-09-20; later batches below, each reviewed on its own.

from datasets import load_dataset

load_dataset("Twu31/bci-report", "results")   # 39 protocol × model scores
load_dataset("Twu31/bci-report", "topics")    # 82 deployment-condition measurements

What is in here

Two separate bodies of measurement. Not one ranking, and rows from one do not belong in a table with rows from the other.

Config	Rows	What it covers	Unit
results	39	9 methods × 8 fixed protocols, 7 public datasets	percent
topics	82	4 deployment questions: dry-vs-wet, on-the-move, calibration-budget, does-pretraining-help	proportion [0,1]
contrasts	18	paired within-participant differences	percentage points
seed_sensitivity	5	repeated training runs of one model	percent
foundation_models	128	16 further foundation-model checkpoints × 8 protocols, frozen probes	percent

Alongside: protocols/ (full descriptor per protocol — preprocessing, split, budget, audit hashes), snapshot.json and deployment-topics.json (the reviewed exports the website itself reads).

Every row in results.csv carries its own license, license_url and attribution, so a row lifted out of that table keeps its credit with it; topics.csv cites its sources in deployment-topics.json under dataset_citations.

evidence-update.json — the 22 September 2026 batch, reviewed under its own manifest: a paired in-ear versus scalp sleep comparison (EESM23, 10 people, identical epochs), four posterior electrodes against all sixteen for eyes open or closed (Alpha Waves, 19 of the source's 20 recordings), a physical-phantom artifact test reporting correlation and predictive R² (dimensionless; R² is negative where the decoder fails and is published as measured), and the adaptation roadmap as it stood that day (planned; the measured results are in adaptation-update.json, below). Heterogeneous by design, so it ships as JSON rather than as a table.

clinical-update.json — the 23 September 2026 batch, reviewed under its own manifest: a 149-person case/control comparison on a Parkinson's disease data set (100 patients, 49 controls, one site) published beside an age-and-sex-only confound comparator, plus three holds that produced no score. Research results, not diagnosis: not diagnostic accuracy, not clinical validation, and no interpretation of any individual. No participant rows, no clinical scores and no per-group demographics are published.

context-update.json — the 27 September 2026 batch: a P300 calibration carried from a PC screen to a VR headset and back (21 people, within-person, two fixed baselines, two timing schemes), and walking-speed classification from dry-electrode EEG printed beside a movement-nuisance comparator (58 people), and a four-person pilot of SSVEP windows accepted when no command was intended — window-level counts, always with coverage beside accuracy. Fixed CPU baselines; no foundation-model or fine-tuning result. One one-person pilot is published as status only.

adaptation-update.json — reviewed 1 October 2026: LaBraM adapted to new people on EEGMAT mental arithmetic (36 people, five participant-disjoint folds) three ways — training only a classification head, the last block as well, or rank-4 LoRA — with the same checkpoint, folds, starting heads, batch order and five-epoch recipe, over three seeds. Cohort means, paired changes with people helped and harmed, per-seed means, trainable parameters and training time. The head-only arm is a short gradient-trained head, so the file points to the core matrix's frozen LaBraM ridge readout on the same folds for scale. One fixed recipe, not a tuned ranking; no memory figures. A next-day experiment on a source under editorial hold is listed as status only, with no figure.

extension-update.json — reviewed 2 October 2026, two fixed classical baselines. The asynchronous SSVEP non-control test extended to the twenty other people of the same release, scored under two rejection rules fixed before scoring: a threshold fitted on the four-person pilot and a personal threshold fitted on 96 of each person's own windows. Detection balanced accuracy, coverage, correct-and-accepted output and false acceptance per non-control state for both, with people helped, harmed and tied. A better detector here is not better command accuracy, and these are window rates, not false activations per hour. And LTRSVP (PhysioNet, doi:10.13026/C2KX0P; Matran-Fernandez and Poli, PLoS ONE 2017, doi:10.1371/journal.pone.0178498; ODC-By 1.0): a P300 target decoder trained on one recording at 5 or 10 Hz and tested on a different recording of the same person at 10 Hz, nine people, with the full 3×3 rate matrix. Within a rate, run b followed run a after a long break; across rates the original study presented the rates from the lowest to the highest, not randomised, and how the released files map onto that sequence is not documented, so a cross-rate cell also differs in elapsed time, fatigue and practice. The paired interval crosses zero, and rate and recording change together: not a causal effect of image rate.

large-source-update.json — reviewed 3 October 2026, two new sources and two separate questions, each with two fixed classical CPU baselines and no shared ranking. Dreem Open Datasets (Guillot et al., IEEE TNSRE 2020, doi:10.1109/TNSRE.2020.3011181; deposit doi:10.5281/zenodo.15900394, MIT as the deposit declares it): five-stage sleep staging against the publisher's consensus in 25 healthy sleepers (DOD-H) and 55 people with obstructive sleep apnoea (DOD-O), two separate experiments, never compared with each other. A training prior that always predicts N2 and a spectral ridge, with accuracy, balanced accuracy, macro F1 and Cohen's kappa, the paired balanced-accuracy gain, and the ridge's per-stage recall, precision and F1. Read accuracy beside balanced accuracy: the ridge reaches about 72% accuracy but 49–57% balanced accuracy and never predicts N1, so its N1 precision is null — not defined, not zero. The prior's balanced accuracy sits slightly above one fifth: a floor, not a chance level. Not clinical, not diagnosis; the recordings' physical units are unresolved, so amplitude-sensitive models are held with no figure. And OpenBMI (Lee et al., GigaScience 2019, doi:10.1093/gigascience/giz002; data doi:10.5524/100542, CC0 1.0): a motor-imagery decoder trained on each person's first session and tested on the final 60 trials of their second, after 0, 10, 20 or 40 labelled trials from it, 51 people. Every mean change carries how many people declined; the relative-PSD gain at 40 trials has an interval that includes zero and is not established. An expanded cohort under the same method, not an independent replication.

reliable-decisions-update.json — reviewed 4 October 2026: route 1 of the decision-research roadmap, when a decoder should decline to decide. No classifier was retrained: the saved test scores of models already published here — spectral ridge, EEGNet and two LaBraM arms on EEGMAT (36 people), standard CCA, CBraMod and EEGNet on BETA (70 people) — rescored under a fixed threshold on calibrated confidence, a coverage target, a certified selective risk (SGR) and a learned reject option, with NLL, ECE and Brier score and a person-specific recalibration whose label cost per new person is counted. Every full-coverage accuracy equals the published one. A fixed threshold accepted 28.4% of BETA trials for one method and 1.5% for another, so compare methods at matched coverage, not at a shared cut-off; how many people had nothing accepted travels with every coverage. The learned reject option erred less than the calibrated confidence for no method. The certified risk accepted nothing on EEGMAT — no error rate, not a zero one — and on BETA some certified folds exceeded their own target: a nominal guarantee across people. Calibration data are participant-disjoint inner cross-fit scores, and rejecting likely errors on known classes is not detecting unfamiliar input. Every contrast carries its paired person-bootstrap interval and a verdict; nothing is ranked where intervals overlap. Every protocol is balanced or uniform by design, so these are method comparisons, never deployment error rates. ds003810 is a crude secondary protocol; the idle protocol (another unit) and a source under editorial hold are left out.

foundation-models-update.json and foundation-models.csv — reviewed 4 October 2026: eleven further EEG foundation models, sixteen encoder checkpoints (REVE Base and Large, LUNA Base and Large, BrainOmni Base, CodeBrain, EEGMamba, ST-EEGFormer Base and Large, four eeg-fm-masking checkpoints, ERP-FM Base, SingLEM and ZUNA 1.1), run as frozen probes on the same eight protocols with the published LaBraM and CBraMod recipe and only the encoder swapped; nine of them also adapted on EEGMAT with the 1 October recipe (a head trained on the frozen encoder against rank-4 LoRA, three seeds; the adaptation rows are in the JSON). foundation-models.csv stacks the site's eight per-protocol CSVs: its first columns are the results config's, in percent, and the rest give each row's status, panel (matrix or masking ablation), chance flag, pretraining exposure, licence and footnote. Kept apart from results, and not a ranking: rows are grouped by family, and one is above or below another only where their 95% intervals do not overlap. Sleep staging is the only protocol where new cells lie entirely above every published row; elsewhere no new cell lies above the published non-foundation row with the highest point estimate, and on BETA no new row lies above standard CCA, which needs no training. Two BrainOmni cells were not run (its tokenizer needs two-second windows): empty, with the reason, never zero. Pretraining exposure is a sourced statement: BETA is in ST-EEGFormer's published pretraining list and TMNRED in SingLEM's; ZUNA 1.1 lists no pretraining data, so its cells are unknown; every other cell, LaBraM's and CBraMod's included, is not in the authors' published pretraining list as checked on 4 October 2026 — not proof that a recording was never seen. Weights licences travel with every row (REVE Responsible Use License, LUNA CC BY-ND 4.0, ERP-FM CC BY-NC-SA 4.0; ZUNA 1.1's model card: research use only, not for diagnosis or clinical use). Timings came from a shared GPU and are not published.

shared-representation-update.json — reviewed 7 October 2026: route 2 of the decision-research roadmap, one representation, several questions. Independent models (A), one shared encoder with a fixed linear head per question (B-lin), a shared hidden layer (B-sh) and that layer conditioned on the question's identity by FiLM (C1), compared at matched data and compute: EEGNet trained from scratch and frozen CBraMod features, three seeds each, and CBraMod adapted by LoRA, one seed and secondary; motor imagery on OpenBMI (51 people) and sleep on BOAS (100 people), with EESM19 (20 people, one seed) as a crude replication. "Fixed heads did as well for less" is supported on neither domain: C1 scored +1.48 pp [+0.09, +2.89] above B-lin on imagery against rest and +6.96 pp [+4.92, +9.09] on five-stage sleep scoring, while B-lin was the cheaper set-up. C1 against B-sh (the same hidden layer without conditioning) was equivalent within the 2 pp margin on both motor-imagery questions and non-inferior on sleep, so the question's identity itself adds nothing measurable. On sleep B-sh carries the lift over B-lin; on motor imagery B-sh against B-lin was inconclusive. On sleep, next-epoch change and time of night were at floor for the fixed heads on EEGNet: reported, not counted. Every contrast carries a difference flag and a margin flag fixed before any result; with 21 primary entries and no multiplicity correction, about one in twenty entries with no true difference may show one by chance. E2 sleep ran after every other result was known (owner decision; its design was fixed at the freeze). Questions are asked by identifier, not in language, and these are method comparisons, never deployment error rates. BOAS is published with three stated gaps: The consent statement does not say whether participants agreed to public sharing or secondary use. The ethics and consent statements come from the publisher's dataset description and README. No peer-reviewed paper describes BOAS. The ethics reference was added to the release in version 1.1.1 (May 2025), and the release does not say when it was granted relative to the recordings. Participants are pseudonymised in the public release, and only cohort aggregates of at least 20 people are published. BOAS: Eduardo López-Larraz, María Sierra-Torralba, Sergio Clemente, Galit Fierro, David Oriol, Javier Minguez, Luis Montesano and Jens G. Klinzing · The Bitbrain Open Access Sleep (BOAS) dataset, OpenNeuro ds005555, version 1.1.3 (2026), doi:10.18112/openneuro.ds005555.v1.1.3. Polysomnography EEG and the human-consensus stage labels only; the headband recordings and the publisher's automatic labels are not used.

later-sessions-update.json: does a decoder trained on a person's earlier session still work later, with no labels from the later session? Motor imagery on WBCIC-SHU from session 1 to session 3 (a relative-spectral ridge baseline and frozen CBraMod features with a ridge readout; the two-class and three-class cohorts are kept apart); RSVP target detection at the publisher's nominal Day 7, Day 80 and Day 200 visits (AUROC with average precision beside it, since targets are about 2.5% of events); and continuous cursor tracking on Forenzo & He, kept as a negative result: the fixed ridge had higher overall error than a constant source-mean comparator on every admitted record, with means, medians and the upper tail all published. Same-person session transfer, offline; not online control and not a ranking.

questions-in-language-update.json and questions-in-language-wordings.json: route 3 of the decision-research roadmap, questions in language. One small head over frozen EEG features, asked each question by a question number, a label template or a natural-language description, on SSVEP (BETA) and sleep (BOAS). Seen questions in their training wording, rewordings and questions the EEG head was never trained on are separate results, never pooled into one zero-shot number. In short: words matched question numbers on seen sleep questions but cost accuracy on SSVEP; most rewordings cost accuracy; questions the head was never trained on were not answered usefully; negation, a boundary probe, was not understood. 35 pre-declared comparisons, no multiplicity correction; the pre-run permutation checks, including two that failed before the owner-approved third passed, are disclosed in the file. The wordings file holds the full wording lists, the held-out partition and the derangements (the site's own text, CC BY 4.0). BOAS is published with three stated gaps: The consent statement does not say whether participants agreed to public sharing or secondary use. The ethics and consent statements come from the publisher's dataset description and README. No peer-reviewed paper describes BOAS. The ethics reference was added to the release in version 1.1.1 (May 2025), and the release does not say when it was granted relative to the recordings. Participants are pseudonymised in the public release. BOAS: Eduardo López-Larraz, María Sierra-Torralba, Sergio Clemente, Galit Fierro, David Oriol, Javier Minguez, Luis Montesano and Jens G. Klinzing · The Bitbrain Open Access Sleep (BOAS) dataset, OpenNeuro ds005555, version 1.1.3 (2026), doi:10.18112/openneuro.ds005555.v1.1.3. Polysomnography EEG and the human-consensus stage labels only; the headband recordings and the publisher's automatic labels are not used.

beyond-scalp-update.json: what can be measured and decoded beyond scalp EEG? Each entry is its own species, signal and metric: none is compared with scalp EEG, none joins the results table, and they share no ranking. Human fNIRS (Bak et al., 30 people): right-hand, left-hand and foot tapping classified by a fixed shrinkage LDA in a person left out of training, against a training prior; executed movement, not motor imagery, offline, and systemic or motion effects are not excluded. One macaque's motor-cortex spiking (MC_RTT, prepared for the Neural Latents Benchmark): finger velocity decoded by a ridge on held-out trials of the same session, against the training mean, on five metrics, including the one on which the ridge is worse. With one animal every figure is that animal's, no interval is reported, and it is not an official Neural Latents Benchmark score. A numerical coverage check of human fNIRS recordings (ds007420, 12 people): counts only, and no classifier score, because the task codes have no verified lookup and the event durations disagree. FALCON H2 (human intracortical handwriting) and Rabut et al. (human functional ultrasound) are prepared, not scored. MC_RTT is animal data, collected under procedures approved by institutional animal care and use committees; the human sources' ethics and consent statements are in the file. Licences, from each repository's record: Bak 2019 fNIRS dataset: CC BY 4.0; MC_RTT: CC BY 4.0; ds007420: CC0-1.0; FALCON H2: CC BY 4.0; Rabut et al. human functional ultrasound through a cranial window: CC BY-NC 4.0. Credit: Sujin Bak, Jinwoo Park, Jaeyoung Shin and Jichai Jeong · Open Access fNIRS Dataset for Classification of Unilateral Finger- and Foot-Tapping, figshare, doi:10.6084/m9.figshare.9783755.v2, CC BY 4.0. Paper: Sujin Bak, Jinwoo Park, Jaeyoung Shin and Jichai Jeong, Open-Access fNIRS Dataset for Classification of Unilateral Finger- and Foot-Tapping, Electronics 8(12), 1486 (2019), doi:10.3390/electronics8121486. · O'Doherty, Joseph (2024) MC_RTT: macaque motor cortex spiking activity during self-paced reaching (Version 0.241017.1444) [Data set]. DANDI archive. https://doi.org/10.48324/dandi.000129/0.241017.1444, CC BY 4.0. Recorded by Joseph E. O'Doherty, Mariana M. B. Cardoso, Joseph G. Makin and Philip N. Sabes in the Sabes Lab at UCSF (Nonhuman Primate Reaching with Multichannel Sensorimotor Cortex Electrophysiology, Zenodo, doi:10.5281/zenodo.3854034; procedure: Makin et al., Journal of Neural Engineering 15(2), 026010 (2018), doi:10.1088/1741-2552/aa9e95). Prepared for the Neural Latents Benchmark: Felix Pei, Joel Ye et al., Neural Latents Benchmark '21: Evaluating latent variable models of neural population activity, NeurIPS 2021 Track on Datasets and Benchmarks, arXiv:2109.04463. · Shibo Zhou, De’Ja Rogers, David A. Boas, Meryem A. Yücel and Yuanyuan Gao · Open-access Multi-Distance fNIRS Dataset Generated During Ball-Squeezing and Motion-Artifact Induction Tasks, OpenNeuro, doi:10.18112/openneuro.ds007420.v1.0.5, CC0. The dataset asks to acknowledge: Gao, Y., Rogers, D., von Lühmann, A., Ortega-Martinez, A., Boas, D.A., & Yücel, M.A. (2023). Short-separation regression incorporated diffuse optical tomography image reconstruction modeling for high-density functional near-infrared spectroscopy. Neurophotonics, 10(2), 025007. https://doi.org/10.1117/1.NPh.10.2.025007. · Fan, Chaofei; Hahn, Nick; Kamdar, Foram; Avansino, Donald; Wilson, Guy; Hochberg, Leigh; Shenoy, Krishna V; Henderson, Jaime; Willett, Frank (2024) FALCON Benchmark H2: Human Handwriting iBCI (Version 0.241029.1403) [Data set]. DANDI archive. https://doi.org/10.48324/dandi.000950/0.241029.1403, CC BY 4.0. Described in: Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication, NeurIPS 2023. Part of the FALCON benchmark. · Claire Rabut, Sumner Norman, Whitney Griggs, Kay Jann, Jonathan Russin, Vasileios Christopoulos, Charles Liu, Richard A. Andersen and Mikhail G. Shapiro · Dataset for: A window to the brain: ultrasound imaging of human neural activity through an acoustically transparent cranial prosthetic, CaltechData, doi:10.22002/f3y3k-em558, CC BY-NC 4.0. Paper: Claire Rabut, Sumner L. Norman, Whitney S. Griggs, Jonathan J. Russin, Kay Jann, Vasileios Christopoulos, Charles Liu, Richard A. Andersen and Mikhail G. Shapiro, Functional ultrasound imaging of human brain activity through an acoustically transparent cranial window, Science Translational Medicine 16, eadj3143 (2024), doi:10.1126/scitranslmed.adj3143.

pilot-studies-update.json: small pilots, outside every leaderboard and model ranking. The Orthosis error-detection dataset (8 people, 83 runs, 2,407 recorded device events: one cohort, counted once) in three configurations: a 64-channel classical EEG baseline, the same baseline on the 58 channels EEGPT supports, and frozen EEGPT with the same supervised probe against those matched classical features, whose paired interval spans zero: neither arm is shown to be better, and the two are not shown to be equivalent. The labels are recorded injected device events, not subjective error recognition, and movement, muscle or device artifacts may contribute; the EEGPT probe is trained on the other people's labels (not zero-shot), and overlap with EEGPT's pretraining data is not excluded. EEGPT's code licence, Apache-2.0, covers the repository’s source code only; its checkpoint licence is unresolved, no original or converted weights are distributed, and its core-matrix cells are in eegpt-core-update.json, below. BINGO Greek and English inner speech (17 subject fields as the source defines them): a fixed spectral baseline near its 15-class prior in every direction, within and across languages, a negative result for this baseline only. The source states no licence; consent is as the project's public deliverable D3.1 (version 3.0) reports it, for its NATO-alphabet recordings in the same laboratory, and no ethics-approval reference is published. The project documents artifact cleaning (ASR, FORCe) on the continuous recordings before trials were cut (D3.1, version 2.0), not re-verified against the file. A mask-aware EEG analysis of a handwriting competition dataset (5 people, four released coded labels): with channels that sit flat at a digital rail masked, relative EEG is just above the training-majority control and above the missingness-mask and mask-plus-time controls, and not separated from elapsed time and trial order; a data-quality and methods case study on a reused development cohort, not handwriting recognition. No ethics-approval or informed-consent statement was found, on the dataset card or in a participating team's challenge paper; the aggregates are published with that gap stated, by the owner's decision. REFED (2 nominal participant folders, 30 stored clips): a computational diagnostic in which source-only controls have smaller errors than EEG, with no interval and no physiological or population claim. Further arms are status only, with no figure. Licences: Orthosis error-detection EEG and EMG dataset: CC BY 4.0; BINGO cross-linguistic inner-speech EEG dataset: No licence stated; EEG-fNIRS-based Handwriting Trajectory Dataset: CC BY-NC-SA 4.0; REFED: real-time dynamic labelled EEG-fNIRS emotion dataset: CC BY-NC-SA 4.0. Credit: Niklas Kueper, Kartik Chari, Judith Bütefür, Julia Habenicht, Tobias Rossol, Su Kyoung Kim, Marc Tabie, Frank Kirchner and Elsa Andrea Kirchner · EEG and EMG dataset for the detection of errors introduced by an active orthosis device (IJCAI’23 CC6 Competition), Zenodo, doi:10.5281/zenodo.8345429, CC BY 4.0. Paper: Niklas Kueper et al., EEG and EMG dataset for the detection of errors introduced by an active orthosis device, Frontiers in Human Neuroscience 18 (2024), doi:10.3389/fnhum.2024.1304311. · BINGO (Brain Imagined-Speech Communication) project, CERTH · cross-linguistic inner-speech EEG dataset, https://bingo-project.gr/datasets/; no licence is stated by the source. Documentation: Fotis P. Kalaganis, Kostas Georgiadis, Maria Kyrou, Spiros Nikolopoulos and Ioannis Kompatsiaris, D3.1 – The BINGO Benchmarking Framework, version 2.0, public deliverable (2025), https://bingo-project.gr/download/397. · lasfk on Hugging Face · EEG-fNIRS-based Handwriting Trajectory Dataset, revision 4036680b5fb0015472d55589da2f34c5a87a3c10, CC BY-NC-SA 4.0 as declared on the dataset card, gated access. · Xiaojun Ning, Jing Wang, Zhiyang Feng, Tianzuo Xin, Shuo Zhang, Shaoqi Zhang, Zheng Lian, Yi Ding, Youfang Lin and Ziyu Jia · REFED: A Subject Real-time Dynamic Labeled EEG-fNIRS Synchronized Recorded Emotion Dataset, NeurIPS 2025 Datasets and Benchmarks Track. Data: REFED2025/REFED-dataset on Hugging Face, revision 0298355c2823946ac5800b2ec12a082bf2dcd01b, CC BY-NC-SA 4.0, gated.

eegpt-core-update.json: EEGPT beside the core matrix, a frozen row like the further foundation encoders', from the same runs as the published LaBraM rows and, except on idle, the CBraMod rows, with only the encoder swapped, published as computed, without a re-run. Its 8 core-matrix cells: balanced accuracy with a participant-bootstrap interval, macro F1 and the chance level on 7 protocols (each percent figure also in the core rows' units), and on the idle protocol (4 people) counts and window metrics with no interval. Its pretraining-attribution rows on mi-rest and EEGMAT, in the fields of deployment-topics.json: the pretrained encoder against the mean of three random initialisations, under a fixed and a train-selected readout, with paired intervals. Kept apart from results, and not a ranking: EEGPT is above or below a published row only where their 95% intervals do not overlap, and several of those gaps are narrow; the file gives each in percentage points. Footnotes travel with the cells: one-second windows ran as four 0.25-second patches with no padding; on BETA with eight channels two channel embeddings were never trained in pretraining; every core window is shorter than EEGPT's 4-second pretraining window; the readout is the site's fixed probe, not the authors' learnable spatial filter and linear head. None of the seven core datasets is in EEGPT's published pretraining list — not proof that a recording was never seen — and BETA's relation to the Tsinghua benchmark that is in it is stated beside the BETA cells. EEGPT's code licence, Apache-2.0, covers the repository’s source code only; its checkpoint licence is unresolved, no original or converted weights are distributed, and no author endorsement is implied. The pilot comparison in pilot-studies-update.json stays outside every leaderboard.

new-ssvep-frequencies-update.json: six follow-up experiments on BETA and one on Wearable-102 asking what lets a decoder answer an SSVEP frequency it never trained on: reference-spectrum queries, matching heads, phase jitter and phase ramps, CCA's mechanism, and a bounded learned correction on top of CCA. data-quality-update.json: Shin2022 checked end to end, a participant-held-out diagnostic with its coverage accounting; eldBETA and ZuCo 2.0 as status only. clinical-iesseeg-update.json: an infant case-control pilot (IESSEEG) with age and recording-setting controls and frozen NeuroLM probes; point estimates, no clinical use. neurolm-eegmat-update.json: NeuroLM's frozen tokenizer features on EEGMAT, a separate panel beside the pilot studies. open-register-20261020.csv: the open register's dataset ladder, 487 rows: every dataset bci.report lists, with the furthest of five verification steps it has reached and the gaps. The full register, with its findings, is at https://bci.report/datasets/register/.

11 of 18 catalogued methods have been scored (9 in results.csv; EEGPT in eegpt-core-update.json and REVE Base in foundation-models-update.json). A catalogued method with no row in any file has not been run, which is not the same as having failed.

Jev-style evaluation

“Jev-style” here means the interface of a decision model like TypeSafe’s Jev, applied to EEG: encode the recording once, then answer several explicit, typed questions about it, each with a probability. The research plan started from a vision paper that applies the same idea to images: Yu & Yao, 2026 · Visual Jev: Accurate and Efficient Decisions from Shared Visual Context (arXiv:2609.25845).

BCI Report asks whether this holds up on public EEG data in three routes, gathered on Jev-style questions on EEG: the evidence:

• Reliable decisions: when a decoder should decline to decide.
• One model, several questions: one shared representation against separate models.
• Questions in language: questions asked in words, including ones the EEG head was never trained on.

Each route's page gives its figures with their intervals, its limits and the release they come from; the third is mostly a negative result for questions the EEG head was never trained on.

What this does not contain

No raw EEG, no per-participant scores, no participant identifiers, no recordings, no embeddings, no model weights. This is a results table, not a data mirror. Reach the recordings through the source column and follow that source's own terms — several of the underlying datasets are redistributable and several are not, and this repository does not relicense any of them.

Read this before ranking anything

• Chance level differs per protocol. 50% for the binary tasks, 2.5% for 40-class BETA SSVEP, 20% for five-stage sleep. A 57% SSVEP result is far above chance; a 57% binary result is barely above it. Sorting the table by primary_percent across protocols compares numbers that do not share a scale.
• Intervals are descriptive, not confidence intervals. They are bootstrap spreads over the scoring set. They do not support significance claims.
• Most comparisons use one seed. Two protocols add a three-seed EEGNet sensitivity check; the main table keeps its original fixed seed.
• Electrode subsets are not headsets. Four- and eight-electrode subsets of laboratory recordings do not validate a physical four- or eight-channel device.
• Pretraining overlap is checked against published lists, not proven absent. For LaBraM and CBraMod, none of the seven datasets is in the authors' published pretraining list (checked 4 October 2026; every source is in foundation-models-update.json), nor for EEGPT (checked 12 October 2026; in eegpt-core-update.json). The released protocol files keep their original sentence, that overlap is unknown unless documented; recording-level overlap was not audited, so a frozen encoder may still have seen related data.

And for topics.csv specifically:

• The unit changes. value is a proportion in [0,1] here, a percent in results.csv. contrasts.csv differences are percentage points. Concatenating the two tables without rescaling silently divides one of them by 100.
• 82 measurements are not 82 studies. They are protocol-specific rows across 4 questions; a question's rows share a cohort and a protocol, so they are not independent evidence.
• Seed spread is not a confidence interval. seed-sensitivity.csv reports what repeated training runs of the same model on the same data did. It describes the optimizer, not the population, and it cannot be read as an error bar on a participant mean.
• These four questions do not extend the 8-protocol matrix. Different protocols, different cohorts, separately reviewed. Keep them apart.

A small cohort is close to its parts

The idle protocol has a cohort of four, and at that size a published mean can be turned back into a count: 1.7% of 60 command trials is one detection, and since two of the four participants selected an always-abstain threshold, that detection belongs to one of the remaining two.

These counts are published anyway, because a false-activation rate of zero produced by people who never activate is the more misleading number. What a reader cannot recover is which person is which — participants are unidentified here and in the public source recordings, and no demographic, session or ordering information is published that would let anyone line them up.

Treat "aggregate" as a description of what is published, not as a claim that it cannot be inverted.

The same holds for the smallest cohorts in the later files. MC_RTT is one macaque, so every figure about it is that animal's own. REFED's errors pool two nominal participant folders, and no per-folder value is published, because one folder's value and the mean would give away the other. The Orthosis pilot has eight people, and no per-person value, minimum or range is published. The handwriting analysis has five people: its intervals are percentiles of resampled five-person means, not one person's score, and its mask-only control's interval is not published, because its bounds would show individual people's scores.

Protocols

Protocol	Dataset	Cohort	Electrodes	Chance	Metric
mi-rest	ds003810	10	15	50.0%	Balanced accuracy
idle	ds005342	4	17	none — see note	Command detection ≤3s
beta-8ch	BETA	70	8	2.5%	Balanced accuracy
beta-4ch	BETA	70	4	2.5%	Balanced accuracy
arithmetic-rest	EEGMAT	36	19	50.0%	Balanced accuracy
p300-target	ds006593	21	19	50.0%	Balanced accuracy
semantic-target	TMNRED / ds005383	30	30	50.0%	Balanced accuracy
sleep-scalp	EESM19 scalp subset	20	6	20.0%	Balanced accuracy

Source datasets and attribution

Results were computed from these public datasets. Credit belongs to their original authors; this repository adds only the measurements.

Dataset	License	Attribution
ds003810	CC0-1.0	Peterson et al. · OpenNeuro ds003810, version 2.0.2. Study: https://pmc.ncbi.nlm.nih.gov/articles/PMC9114495/
EEGMAT	Open Data Commons Attribution License 1.0	Igor Zyma, Ivan Seleznov, Anton Popov, Mariia Chernykh, Oleksii Shpenkov · EEG During Mental Arithmetic Tasks 1.0.0, PhysioNet, doi:10.13026/C2JQ1P. Study: Zyma et al. (2019), doi:10.3390/data4010014. PhysioNet platform: Pollard et al. (2026), doi:10.1038/s44360-026-00096-z.
EESM19 scalp subset	CC0-1.0 declared by upstream and mirror	Kaare B. Mikkelsen et al. · Accurate whole-night sleep monitoring with dry-contact ear-EEG (2019), doi:10.1038/s41598-019-53115-3; OpenNeuro ds005185 v1.0.2. Processed mirror: Zachary1150/EESM19-Processed.
BETA	CC BY 4.0	Bingchuan Liu et al. · BETA: A Large Benchmark Database Toward SSVEP-BCI Application (2020), doi:10.3389/fnins.2020.00627. Figshare 12264401 v3; mirror Bingchuan/BETA.
TMNRED / ds005383	OpenNeuro metadata says CC0; accompanying publication/GitHub says CC BY 4.0	Yanru Bai, Qi Tang et al. · TMNRED, A Chinese Language EEG Dataset for Fuzzy Semantic Target Identification in Natural Reading Environments (2025), doi:10.1038/s41597-025-05036-2. OpenNeuro ds005383 v1.0.0.
ds006593	CC0-1.0	OpenNeuro ds006593 contributors · version 1.0.0, doi:10.18112/openneuro.ds006593.v1.0.0; original author credits retained at the linked source.
ds005342	CC0-1.0	OpenNeuro ds005342 contributors · version 1.0.3, doi:10.18112/openneuro.ds005342.v1.0.3; associated study doi:10.3389/fninf.2022.961089. Original author credits are retained at the linked source.

Corrections to released wording

Released files are fixed bytes, so wording errors are corrected here and on the site rather than in place (full list: https://bci.report/releases/#corrections).

• deployment-topics.json calls the original ensemble-TRCA method a three-filter-bank experiment. The original paper and the reference code's tutorials use five sub-bands; three is only the default of the code's functions. No score changes.
• For ds003810, the OpenNeuro record asks users to cite Peterson, Galván, Hernández and Spies, A feasibility study of a complete low-cost consumer-grade brain-computer interface system, Heliyon 6(3):e03425 (2020), doi:10.1016/j.heliyon.2020.e03425 — in addition to the Data in Brief description linked in the rows.

Citation

Cite BCI Report and the release you used — this card was built at release open-register-update-20261020 (2026-10-20) — and the upstream dataset each figure was computed on: the attribution table above, and every dataset page at https://bci.report/datasets/, gives its credit. The same citation is in CITATION.cff, which GitHub's "Cite this repository" reads, and every topic, dataset and method page on the site ends with a "Cite this page" block naming the releases its figures come from. Every release is archived on Zenodo; the concept DOI 10.5281/zenodo.23123296 resolves to the newest one.

@misc{bcireport,
  title        = {BCI Report: public EEG decoding results reported with their protocol},
  author       = {{BCI Report}},
  year         = {2026},
  howpublished = {\url{https://bci.report}},
  doi          = {10.5281/zenodo.23123296},
  note         = {Release open-register-update-20261020}
}

License

The cc-by-4.0 tag covers the aggregate result tables and protocol descriptors in this repository — measurements this project produced. It does not and cannot relicense the underlying recordings, whose terms are listed per row and per dataset above. The code that produced them, on GitHub, is under the MIT License.

Corrections

Scientific corrections, benchmark-method questions, and rights, privacy or attribution concerns are welcome, including ones that would withdraw a published result — affected results can be withheld while an issue is reviewed. Open a discussion here, or write to contact@bci.report for scientific corrections and privacy@bci.report for rights, privacy and withdrawal requests. Please do not send raw EEG, participant names or health records.

Full source review, privacy reasoning and interpretation caveats: https://bci.report/data-use/