{ "generatedAt": "2026-09-20", "releaseId": "research-preview-20260920", "tracks": [ { "id": "mi-rest", "title": "Motor imagery & rest", "short": "Transfer to a new person", "dataset": "ds003810", "subtitle": "Rest versus right-hand imagery", "type": "accuracy", "subjects": 10, "observations": "1,200 epochs · 5 participant-disjoint folds", "exposure": "15 channels · 2-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Ten-person laboratory task with prompted rest. This is not continuous-idle monitoring or a physical low-channel headset test.", "protocol": [ "5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.", "epoching: two seconds from the class annotation onset; amplitude: converted to microvolts according to source calibration, then per-channel epoch mean removed; resampling: none; native sampling rate retained; selection: natural file/event order; when over the cap, retain 60 evenly spaced event indices per participant/class; source_units: Microv declared by BIDS channels.tsv; EDF physical dimension is absent, so MNE returns the source numeric microvolt values without SI scaling", "One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 20 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).", "Labels: rest, right_hand_imagery.", "Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.", "run 0 is real dominant-hand movement and is excluded; only imagery runs 1-4 contribute", "the source reports online 0.5-45 Hz filtering", "controlled cue-locked laboratory windows; this does not measure continuous false activations", "foundation-model pretraining overlap is unknown", "EEGNet three-seed mean 69.47%; sample SD 1.43 percentage points; range 68.17–71.00%. Main table retains the original fixed seed; this is not a confidence interval." ], "protocolId": "parallel-fixed-subject-folds-v1/ds003810", "version": "OpenNeuro snapshot 2.0.2 · c674303319303b9ffc4ff7e2340b3a4807e8ffb5", "elapsed": 71.65976849989966, "peakGb": null, "auditSha": "e762cf7f048bf2df47b9399b9df6171733a93c162ce7800560952df30491a7f2", "source": "https://doi.org/10.18112/openneuro.ds003810.v2.0.2", "backend": "Apple M5 / MPS", "chanceLevel": 50.0, "selection": "EEGNet three-seed mean 69.47%; sample SD 1.43 percentage points; range 68.17–71.00%. Main table retains the original fixed seed; this is not a confidence interval.", "rows": [ { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 15, "mode": "Supervised fit", "x": 0.5440653297848919, "y": 54.91666666666667, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 53.2–56.7%", "interval": [ 53.16666666666665, 56.666666666666664 ], "abstain": null, "seconds": 0.17833587527275085, "subjects": 10, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "csp-lda", "name": "CSP+LDA", "family": "classical", "parameters": null, "channels": 15, "mode": "Supervised fit", "x": 0.4055403110906564, "y": 47.08333333333332, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 43.8–50.2%", "interval": [ 43.75, 50.249999999999986 ], "abstain": null, "seconds": 8.815756499767303, "subjects": 10, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 15, "mode": "Frozen encoder + ridge head", "x": 0.5093637206233653, "y": 53.5, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 51.4–55.4%", "interval": [ 51.41666666666668, 55.41666666666668 ], "abstain": null, "seconds": 2.1763478331267834, "subjects": 10, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 15, "mode": "Frozen encoder + ridge head", "x": 0.6173268662572962, "y": 62.916666666666664, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 59.6–66.5%", "interval": [ 59.58333333333334, 66.5 ], "abstain": null, "seconds": 4.9904160420410335, "subjects": 10, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 15, "mode": "Scratch · 20 epochs", "x": 0.6799956127418472, "y": 71.0, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 65.0–76.3%", "interval": [ 64.99791666666667, 76.33333333333333 ], "abstain": null, "seconds": 55.498912249691784, "subjects": 10, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed. EEGNet three-seed mean 69.47%; sample SD 1.43 percentage points; range 68.17–71.00%. Main table retains the original fixed seed; this is not a confidence interval.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "summarySha": "0dd66f4bd4d083fca1d1ab44a2d2951d57f6d19378079b8cb5ff3ee9241f0cda", "protocolSha": "57ddb7692e3bb248b95c4e81b6aa02e7c8b19d774ed4a8b799c811b6ef3b65a2", "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "attribution": "Peterson et al. · OpenNeuro ds003810, version 2.0.2. Study: https://pmc.ncbi.nlm.nih.gov/articles/PMC9114495/", "privacyReview": "Study paper reports ethics approval by CEySTE CCT-CONICET Santa Fe, Declaration of Helsinki compliance, and signed consent. Twelve adults participated; ten were retained. Public participants.tsv includes exact age, sex, handedness, and pseudonymous ID, which must not be surfaced.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://doi.org/10.18112/openneuro.ds003810.v2.0.2", "https://pmc.ncbi.nlm.nih.gov/articles/PMC9114495/" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim.", "seedSensitivity": { "model": "EEGNet", "seeds": [ 20260919, 20260920, 20260921 ], "balancedAccuracyPercent": [ 71.0, 69.25, 68.16666666666666 ], "meanPercent": 69.47222222222221, "sampleSdPercentagePoints": 1.4296788036064232, "rangePercentagePoints": 2.833333333333343, "auditSha": "17b31f77610fcb5f034eb013774b5f796b96b3f393ec763351f1c4c290e5c194", "scope": "Same participants, folds, preprocessing and 20-epoch budget. Three seeds measure initialization variability, not population uncertainty. Main table retains its preselected seed; no best-seed selection." } }, { "id": "idle", "title": "Idle & command", "short": "False activation × detection", "dataset": "ds005342", "subtitle": "Seated motor imagery · cue-gated replay", "type": "tradeoff", "subjects": 4, "observations": "60 idle / 60 command test trials", "exposure": "180 seconds of test idle", "status": "Research preview", "xLabel": "Idle false activation", "yLabel": "Command detection ≤3s", "limitation": "Detection resets at each cue. Only three minutes of test idle are observed; these results cannot estimate natural, always-on false activations per hour.", "protocol": [ "For each subject, the first two blocks train the model, the third selects the threshold, and the final block is held out for testing.", "Class events 1/2 define task onset. Two-second history windows produce decisions at 2.0, 2.5 and 3.0 seconds.", "Two consecutive threshold crossings trigger one command. Thresholds use calibration data only and may reject every command.", "17 channels at 250 Hz; one-way 4–40 Hz filtering. Normalization uses training data only; no future samples enter a decision window.", "Visual cues, original feedback and short calibration exposure can influence results. This is not a natural continuous-idle test.", "Additional models use protocol ds005342-sitting-cue-gated-expanded-mvp-v1; independent audit SHA256: 14445d3423b5268654a1f73f7e1aca8bb03a2f793b076f450883fea89c7bcc83." ], "protocolId": "ds005342-sitting-cue-gated-pilot-v1", "version": "OpenNeuro snapshot 1.0.3 · 30bb3b67d76ce007b822fed801e496fafe5598a0", "elapsed": 57.95052358601242, "peakGb": 2.364063744, "auditSha": "ce2474c547f5f11c256752b0a07466beb5b6928da34bb39783aaa4647973d2e4", "source": "https://doi.org/10.18112/openneuro.ds005342.v1.0.3", "rows": [ { "id": "csp-lda", "name": "CSP+LDA", "family": "classical", "parameters": null, "channels": 17, "mode": "Supervised fit", "x": 0.0, "y": 23.333333333333332, "xDetail": "0 / 60 idle trials", "yDetail": "14 / 60 command trials", "abstain": 1, "seconds": 0.1350538763217628, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": 5819936, "channels": 17, "mode": "Frozen encoder + linear head", "x": 3.3333333333333335, "y": 18.333333333333332, "xDetail": "2 / 60 idle trials", "yDetail": "11 / 60 command trials", "abstain": 0, "seconds": 2.84153566788882, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": 1858, "channels": 17, "mode": "Scratch · 20 epochs", "x": 0.0, "y": 1.6666666666666667, "xDetail": "0 / 60 idle trials", "yDetail": "1 / 60 command trials", "abstain": 2, "seconds": 10.115781626664102, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "shallowfbcspnet", "name": "ShallowFBCSPNet", "family": "small", "parameters": 30482, "channels": 17, "mode": "Scratch · 20 epochs", "x": 1.6666666666666667, "y": 43.333333333333336, "xDetail": "1 / 60 idle trials", "yDetail": "26 / 60 command trials", "abstain": 1, "seconds": 28.6309855398722, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": 4883800, "channels": 17, "mode": "Frozen encoder + linear head", "x": 1.6666666666666667, "y": 18.333333333333332, "xDetail": "1 / 60 idle trials", "yDetail": "11 / 60 command trials", "abstain": 1, "seconds": 2.349851499311626, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "deep4net", "name": "Deep4Net", "family": "small", "parameters": 274552, "channels": 17, "mode": "Scratch · 20 epochs", "x": 1.6666666666666667, "y": 3.3333333333333335, "xDetail": "1 / 60 idle trials", "yDetail": "2 / 60 command trials", "abstain": 1, "seconds": 13.877315375953913, "subjects": 4, "note": "Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "attribution": "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.", "privacyReview": "README reports voluntary signed consent and Universidad Antonio Nariño ethics approval. participants.tsv links pseudonymous IDs to exact age, sex, handedness, and academic major. The current evaluated subset has only four people.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://doi.org/10.18112/openneuro.ds005342.v1.0.3", "https://eegdash.org/api/dataset/eegdash.dataset.DS005342.html", "https://doi.org/10.3389/fninf.2022.961089" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim.", "backend": "Apple M5 / MPS and CPU", "chanceLevel": null, "selection": "Single fixed configuration; no test-set hyperparameter selection." }, { "id": "beta-8ch", "title": "SSVEP · 8 channels", "short": "Transfer to a new person", "dataset": "BETA", "subtitle": "40 visual targets · 8 posterior electrodes", "type": "accuracy", "subjects": 70, "observations": "11,200 trials · 7 participant-disjoint folds", "exposure": "8 channels selected from a 64-channel recording · 2-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Near-floor scores are not ordered reliably between the 8- and 4-electrode subsets. Electrode subsets from laboratory recordings do not validate a physical low-channel cap. Prompted SSVEP does not measure idle false activations. Single-seed results; pretraining overlap unknown.", "protocol": [ "Seven participant-disjoint folds: train on 60 people, test on ten. All four blocks stay with their participant.", "Two seconds from stimulus onset; no visual-latency shift. Source data were already zero-phase filtered. Additional 6–80 Hz filtering applies to each selected window separately.", "Microvolt units are inferred from an independently documented loader, not explicitly stated in the author MAT description. Inconsistent phase metadata are unused by all methods.", "Standard CCA uses known frequencies and three harmonics without training labels. Frozen encoders use training-only standardized ridge heads (alpha 100). EEGNet trains from scratch for 20 epochs with one seed (20260912).", "Electrodes: PO5, PO3, POZ, PO4, PO6, O1, OZ, O2.", "Uniform-guessing reference: 2.5%. Descriptive 95% intervals resample participants; training sets overlap across folds.", "No cross-task overall ranking, model fine-tuning optimum or hardware benchmark is claimed." ], "protocolId": "beta-ssvep-2s-posterior4and8-subject7fold-v1/posterior8", "version": "Hugging Face mirror of BETA database, Figshare record 12264401 v3 · {\"mirror\": \"d4290c0200db8a104e0f557349dc49f90ba79506\", \"upstream\": \"Figshare version 3, 2022-06-15\"}", "elapsed": 298.7745355004445, "peakGb": null, "auditSha": "ccddf911864b8471f488e46cae45a3651369dae1134afcdeaede68bdfcc31950", "summarySha": "547a35763f7bca10a65c7b44c2f8c46af290159c692d237e92cd682fc21127a6", "protocolSha": "dcd1d2bf4dbb46b7db8c4562623448633cde4641fa372c90f25ba9d84012c2f0", "source": "https://figshare.com/articles/dataset/The_BETA_database/12264401", "backend": "Apple M5 / MPS and CPU", "chanceLevel": 2.5, "selection": "One fixed seed and training budget; multi-seed sensitivity pending.", "rows": [ { "id": "cca", "name": "Standard CCA", "family": "classical", "parameters": null, "channels": 8, "mode": "Known-frequency reference · no fitting", "x": 0.6262284336377374, "y": 63.11607142857143, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 57.2–69.0%", "interval": [ 57.24062499999999, 68.96428571428572 ], "abstain": null, "seconds": 1.754876459017396, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 8, "mode": "Supervised fit", "x": 0.49373413893529433, "y": 50.705357142857146, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 45.6–55.8%", "interval": [ 45.56227678571428, 55.758928571428555 ], "abstain": null, "seconds": 1.366779665928334, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 8, "mode": "Frozen encoder + ridge head", "x": 0.09296091863206704, "y": 10.830357142857142, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 9.5–12.2%", "interval": [ 9.526562499999999, 12.196651785714284 ], "abstain": null, "seconds": 5.000138541217893, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 8, "mode": "Frozen encoder + ridge head", "x": 0.31778213986710313, "y": 33.66964285714286, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 29.8–37.6%", "interval": [ 29.83928571428572, 37.60714285714286 ], "abstain": null, "seconds": 4.85779504198581, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 8, "mode": "Scratch · 20 epochs", "x": 0.541059342838019, "y": 55.75892857142857, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 50.1–61.3%", "interval": [ 50.14285714285713, 61.2955357142857 ], "abstain": null, "seconds": 285.79494579229504, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "license": "CC BY 4.0", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "attribution": "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.", "privacyReview": "Paper reports Tsinghua University ethics approval 20190002 and written consent; parents consented for participants under 16. Ages span 9-64. Only cohort aggregates should be shown.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://figshare.com/articles/dataset/The_BETA_database/12264401", "https://doi.org/10.3389/fnins.2020.00627" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim." }, { "id": "beta-4ch", "title": "SSVEP · 4 channels", "short": "Transfer to a new person", "dataset": "BETA", "subtitle": "40 visual targets · 4 posterior electrodes", "type": "accuracy", "subjects": 70, "observations": "11,200 trials · 7 participant-disjoint folds", "exposure": "4 channels selected from a 64-channel recording · 2-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Near-floor scores are not ordered reliably between the 8- and 4-electrode subsets. Electrode subsets from laboratory recordings do not validate a physical low-channel cap. Prompted SSVEP does not measure idle false activations. Single-seed results; pretraining overlap unknown.", "protocol": [ "Seven participant-disjoint folds: train on 60 people, test on ten. All four blocks stay with their participant.", "Two seconds from stimulus onset; no visual-latency shift. Source data were already zero-phase filtered. Additional 6–80 Hz filtering applies to each selected window separately.", "Microvolt units are inferred from an independently documented loader, not explicitly stated in the author MAT description. Inconsistent phase metadata are unused by all methods.", "Standard CCA uses known frequencies and three harmonics without training labels. Frozen encoders use training-only standardized ridge heads (alpha 100). EEGNet trains from scratch for 20 epochs with one seed (20260912).", "Electrodes: POZ, O1, OZ, O2.", "Uniform-guessing reference: 2.5%. Descriptive 95% intervals resample participants; training sets overlap across folds.", "No cross-task overall ranking, model fine-tuning optimum or hardware benchmark is claimed." ], "protocolId": "beta-ssvep-2s-posterior4and8-subject7fold-v1/posterior4", "version": "Hugging Face mirror of BETA database, Figshare record 12264401 v3 · {\"mirror\": \"d4290c0200db8a104e0f557349dc49f90ba79506\", \"upstream\": \"Figshare version 3, 2022-06-15\"}", "elapsed": 153.87726529221982, "peakGb": null, "auditSha": "ccddf911864b8471f488e46cae45a3651369dae1134afcdeaede68bdfcc31950", "summarySha": "547a35763f7bca10a65c7b44c2f8c46af290159c692d237e92cd682fc21127a6", "protocolSha": "dcd1d2bf4dbb46b7db8c4562623448633cde4641fa372c90f25ba9d84012c2f0", "source": "https://figshare.com/articles/dataset/The_BETA_database/12264401", "backend": "Apple M5 / MPS and CPU", "chanceLevel": 2.5, "selection": "One fixed seed and training budget; multi-seed sensitivity pending.", "rows": [ { "id": "cca", "name": "Standard CCA", "family": "classical", "parameters": null, "channels": 4, "mode": "Known-frequency reference · no fitting", "x": 0.5670723362358747, "y": 57.58035714285715, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 51.8–63.5%", "interval": [ 51.79464285714286, 63.4825892857143 ], "abstain": null, "seconds": 1.086679874919355, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 4, "mode": "Supervised fit", "x": 0.4671589176778854, "y": 48.214285714285715, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 43.1–53.3%", "interval": [ 43.0625, 53.29486607142857 ], "abstain": null, "seconds": 0.7049391251057386, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 4, "mode": "Frozen encoder + ridge head", "x": 0.110318056846, "y": 12.883928571428571, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 11.2–14.6%", "interval": [ 11.23214285714286, 14.634151785714286 ], "abstain": null, "seconds": 2.451408333145082, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 4, "mode": "Frozen encoder + ridge head", "x": 0.2568676299952549, "y": 27.607142857142858, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 24.1–31.3%", "interval": [ 24.080357142857135, 31.277008928571426 ], "abstain": null, "seconds": 2.954518625047058, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 4, "mode": "Scratch · 20 epochs", "x": 0.42073851241459875, "y": 44.125, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 38.6–49.6%", "interval": [ 38.642857142857146, 49.63392857142857 ], "abstain": null, "seconds": 146.67971933400258, "subjects": 70, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "license": "CC BY 4.0", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "attribution": "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.", "privacyReview": "Paper reports Tsinghua University ethics approval 20190002 and written consent; parents consented for participants under 16. Ages span 9-64. Only cohort aggregates should be shown.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://figshare.com/articles/dataset/The_BETA_database/12264401", "https://doi.org/10.3389/fnins.2020.00627" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim." }, { "id": "arithmetic-rest", "title": "Arithmetic & rest", "short": "Transfer to a new person", "dataset": "EEGMAT", "subtitle": "Serial subtraction versus resting EEG", "type": "accuracy", "subjects": 36, "observations": "2,160 epochs · 5 participant-disjoint folds", "exposure": "19 channels · 2-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Thirty nonoverlapping 2-second windows from each of two conditions per person. Published signals were already cleaned with ICA; task performance groups are not evaluated.", "protocol": [ "5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.", "First 60 s of each recording; 30 contiguous nonoverlapping 2 s windows; exclude A2-A1 ear-difference and ECG; EDF physical volts converted to microvolts; subtract each channel's window mean; no rejection, filtering, resampling, or learned preprocessing.", "One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 20 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).", "Labels: pre-task-rest, mental-arithmetic.", "Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.", "The benchmark detects condition (rest versus serial subtraction), not the good/bad count-quality participant grouping.", "Only the first documented 60 seconds is retained even though EDF containers are longer.", "The source README reports prior ICA artifact removal, so these are not untouched acquisition signals.", "Open Data Commons Attribution License v1.0 applies; retain PhysioNet attribution.", "EEGNet three-seed mean 67.52%; sample SD 0.14 percentage points; range 67.36–67.64%. Main table retains the original fixed seed; this is not a confidence interval." ], "protocolId": "parallel-fixed-subject-folds-v1/physionet-eegmat-1.0.0", "version": "PhysioNet EEG During Mental Arithmetic Tasks 1.0.0 · 1.0.0", "elapsed": 126.49400083348155, "peakGb": null, "auditSha": "2d0e4cec78c8d6e5bfc9f5390ba161f4308759e02feffbb8645d6983d9093292", "source": "https://physionet.org/content/eegmat/1.0.0/", "backend": "Apple M5 / MPS", "chanceLevel": 50.0, "selection": "EEGNet three-seed mean 67.52%; sample SD 0.14 percentage points; range 67.36–67.64%. Main table retains the original fixed seed; this is not a confidence interval.", "rows": [ { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 19, "mode": "Supervised fit", "x": 0.5576696101150709, "y": 56.75925925925925, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 54.6–58.9%", "interval": [ 54.58333333333333, 58.93518518518519 ], "abstain": null, "seconds": 0.5772068751975894, "subjects": 36, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 19, "mode": "Frozen encoder + ridge head", "x": 0.6197865063617024, "y": 64.62962962962963, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 60.5–68.6%", "interval": [ 60.50925925925926, 68.56481481481481 ], "abstain": null, "seconds": 2.1183615000918508, "subjects": 36, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 19, "mode": "Frozen encoder + ridge head", "x": 0.6020541673231031, "y": 62.31481481481481, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 58.1–66.5%", "interval": [ 58.10069444444444, 66.48148148148148 ], "abstain": null, "seconds": 6.100336125120521, "subjects": 36, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 19, "mode": "Scratch · 20 epochs", "x": 0.6367413737850042, "y": 67.63888888888889, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 62.8–72.5%", "interval": [ 62.824074074074076, 72.45370370370368 ], "abstain": null, "seconds": 117.69809633307159, "subjects": 36, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed. EEGNet three-seed mean 67.52%; sample SD 0.14 percentage points; range 67.36–67.64%. Main table retains the original fixed seed; this is not a confidence interval.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "summarySha": "ab607e749da1eedcf88441c5b6fd0db55e2ebae0bbd91d6d18e508605eef625b", "protocolSha": "5d53ccbd5a00763908162f0b700c0a3236e0d4f9d43b08e42ee31a2ef01aefbf", "license": "Open Data Commons Attribution License 1.0", "licenseUrl": "https://opendatacommons.org/licenses/by/1-0/", "attribution": "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.", "privacyReview": "Paper reports Bioethics Commission approval (15 August 2018) and written consent from every participant. EDF dates are normalized, but subject-info.csv contains age, gender, occupation, and recording date; those fields must not be published.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://physionet.org/content/eegmat/1.0.0/", "https://www.mdpi.com/2306-5729/4/1/14", "https://opendatacommons.org/licenses/by/1-0/" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim.", "seedSensitivity": { "model": "EEGNet", "seeds": [ 20260919, 20260920, 20260921 ], "balancedAccuracyPercent": [ 67.63888888888889, 67.5462962962963, 67.3611111111111 ], "meanPercent": 67.51543209876543, "sampleSdPercentagePoints": 0.1414375214492599, "rangePercentagePoints": 0.2777777777777857, "auditSha": "17b31f77610fcb5f034eb013774b5f796b96b3f393ec763351f1c4c290e5c194", "scope": "Same participants, folds, preprocessing and 20-epoch budget. Three seeds measure initialization variability, not population uncertainty. Main table retains its preselected seed; no best-seed selection." } }, { "id": "p300-target", "title": "P300 target ERP", "short": "Transfer to a new person", "dataset": "ds006593", "subtitle": "Visual target versus nontarget events", "type": "accuracy", "subjects": 21, "observations": "2,520 epochs · 5 participant-disjoint folds", "exposure": "19 channels · 1-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Balanced target/nontarget sample changes the source 1:9 prevalence. One-second windows include later flashes at 0.3-second intervals; this is not an online speller estimate.", "protocol": [ "5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.", "units: microvolts; epoch: [stimulus onset, onset + 1.0 s); filtering: none added", "One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 20 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).", "Labels: nontarget, target.", "Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.", "Research subset with fixed deterministic sampling capped at 60 epochs per participant/class.", "The cap balances target and nontarget and therefore changes the original approximately 1:9 class prevalence.", "Stimuli were presented about every 0.3 s, so each 1 s epoch contains responses to later stimuli; this is intrinsic next-stimulus contamination." ], "protocolId": "parallel-fixed-subject-folds-v1/ds006593", "version": "OpenNeuro snapshot 1.0.0 · b3fa345b310abfcf442dd4e2867f6652336c3fe4", "elapsed": 61.87171109719202, "peakGb": null, "auditSha": "30762120c57bede9f47d2e96637656e8948fca458016f7f2e3a39f951a357716", "source": "https://doi.org/10.18112/openneuro.ds006593.v1.0.0", "backend": "Local Ubuntu / CUDA", "chanceLevel": 50.0, "selection": "One fixed seed and training budget; multi-seed sensitivity pending.", "rows": [ { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 19, "mode": "Supervised fit", "x": 0.5051245210147807, "y": 50.95238095238096, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 49.6–52.4%", "interval": [ 49.56349206349206, 52.38095238095239 ], "abstain": null, "seconds": 0.3815236510708928, "subjects": 21, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "temporal-ridge", "name": "Temporal ridge", "family": "classical", "parameters": null, "channels": 19, "mode": "Supervised fit", "x": 0.5201894836660051, "y": 53.80952380952382, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 51.9–55.6%", "interval": [ 51.94444444444445, 55.635912698412696 ], "abstain": null, "seconds": 0.4925835959147662, "subjects": 21, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 19, "mode": "Frozen encoder + ridge head", "x": 0.48578804304247825, "y": 49.4047619047619, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 47.7–51.2%", "interval": [ 47.65873015873016, 51.19047619047619 ], "abstain": null, "seconds": 10.641786107094958, "subjects": 21, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 19, "mode": "Frozen encoder + ridge head", "x": 0.5290608291102008, "y": 54.960317460317455, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 53.3–56.7%", "interval": [ 53.25396825396827, 56.706349206349216 ], "abstain": null, "seconds": 4.786083345999941, "subjects": 21, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 19, "mode": "Scratch · 20 epochs", "x": 0.5003107165141089, "y": 53.293650793650805, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 51.7–54.9%", "interval": [ 51.70634920634922, 54.92162698412698 ], "abstain": null, "seconds": 45.56973439711146, "subjects": 21, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "summarySha": "a63a7ba5bbd6dd7865ae260b707c2db896136646a910d96516fa5c862e1146c0", "protocolSha": "bf6796dfe04dffa57bdd08933d2bb743722da71594135f04fa4fc33f3fb7dd46", "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "attribution": "OpenNeuro ds006593 contributors · version 1.0.0, doi:10.18112/openneuro.ds006593.v1.0.0; original author credits retained at the linked source.", "privacyReview": "Dataset metadata/README report IRB approval #10-05-2022 and written informed consent. Participant metadata contains pseudonymous IDs and study descriptors without demographic fields. Eye-tracking and raw signals remain off-site.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://doi.org/10.18112/openneuro.ds006593.v1.0.0", "https://nemar.org/dataset/on006593" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim." }, { "id": "semantic-target", "title": "Semantic target ERP", "short": "Transfer to a new person", "dataset": "TMNRED / ds005383", "subtitle": "Reading · target versus nontarget events", "type": "accuracy", "subjects": 30, "observations": "3,600 epochs · 5 participant-disjoint folds", "exposure": "30 channels · 1-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Balanced event subset with collapsed semantic categories. Natural class prevalence is not preserved. Third-party reading passages and participant metadata are excluded from this website.", "protocol": [ "5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.", "units: microvolts; epoch: [event onset, onset + 1.0 s); filtering: none added", "One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 20 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).", "Labels: nontarget, target.", "Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.", "Research subset with fixed deterministic sampling capped at 60 epochs per participant/class.", "Balancing to 60 per class changes the source target/nontarget prevalence.", "Events use numbered target/nontarget variants; the benchmark collapses the suffix only because the trial_type prefix explicitly names the semantic class." ], "protocolId": "parallel-fixed-subject-folds-v1/ds005383", "version": "OpenNeuro snapshot 1.0.0 · ad78f3db430e636595b1b1c08417492f3067a25e", "elapsed": 112.4083747221157, "peakGb": null, "auditSha": "5920974ae9ca3876ce1be82497c139f0783d2ff47173641e329af976441de4c4", "source": "https://doi.org/10.18112/openneuro.ds005383.v1.0.0", "backend": "Local Ubuntu / CUDA", "chanceLevel": 50.0, "selection": "One fixed seed and training budget; multi-seed sensitivity pending.", "rows": [ { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 30, "mode": "Supervised fit", "x": 0.506020854240093, "y": 51.027777777777786, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 49.6–52.5%", "interval": [ 49.55555555555556, 52.52777777777777 ], "abstain": null, "seconds": 0.8656524890102446, "subjects": 30, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "temporal-ridge", "name": "Temporal ridge", "family": "classical", "parameters": null, "channels": 30, "mode": "Supervised fit", "x": 0.5585273139742724, "y": 56.33333333333332, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 54.4–58.4%", "interval": [ 54.41666666666667, 58.36111111111111 ], "abstain": null, "seconds": 1.2127933809533715, "subjects": 30, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 30, "mode": "Frozen encoder + ridge head", "x": 0.5188597352835342, "y": 53.166666666666664, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 51.3–54.9%", "interval": [ 51.27777777777778, 54.94444444444445 ], "abstain": null, "seconds": 3.368051374098286, "subjects": 30, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 30, "mode": "Frozen encoder + ridge head", "x": 0.5536632458905991, "y": 55.80555555555555, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 54.2–57.4%", "interval": [ 54.19444444444444, 57.44444444444443 ], "abstain": null, "seconds": 23.52429261500947, "subjects": 30, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 30, "mode": "Scratch · 20 epochs", "x": 0.6058092006566062, "y": 61.44444444444444, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 58.9–64.0%", "interval": [ 58.916666666666664, 63.97222222222221 ], "abstain": null, "seconds": 83.43758486304432, "subjects": 30, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "summarySha": "c5dca7fc30d07a001b8980a4ea65e55ef292f0891a4d32a12262b8ec88389243", "protocolSha": "839dfba89cf200aac7a28a7259f45024d006fcd4500170e247fcc1c4e9962c8d", "license": "OpenNeuro metadata says CC0; accompanying publication/GitHub says CC BY 4.0", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "attribution": "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.", "privacyReview": "Paper reports Tianjin University ethics approval TJUE-2024-402, written consent, deidentification, and explicit participant permission for open sharing. Local participants.tsv appears column-shifted/malformed, so it must not feed public metadata.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://doi.org/10.18112/openneuro.ds005383.v1.0.0", "https://www.nature.com/articles/s41597-025-05036-2" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim." }, { "id": "sleep-scalp", "title": "Sleep staging", "short": "Transfer to a new person", "dataset": "EESM19 scalp subset", "subtitle": "Five stages · balanced scalp-EEG sample", "type": "accuracy", "subjects": 20, "observations": "3,000 epochs · 5 participant-disjoint folds", "exposure": "6 channels · 30-second windows", "status": "Research preview", "xLabel": "Macro F1", "yLabel": "Balanced accuracy", "limitation": "Balanced quality-screened scalp subset, not whole-night deployment prevalence and not ear-EEG. Frozen encoders average fifteen 2-second representations; EEGNet receives 10 epochs.", "protocol": [ "5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.", "Retain six named scalp electrodes; exclude mastoids. Drop any epoch with source per-channel missing-value flag in those six electrodes, actual nonfinite values, any channel std<0.01uV, or peak-to-peak>1000uV. Use full30s at200Hz, no further filtering/reference/amplitude transformation. Select up to30 evenly spaced eligible epochs per participant/class across available nights.", "One fixed seed (20260919); no early stopping or test-based tuning. EEGNet trains for 10 epochs per fold. Frozen encoders use training-only standardized ridge heads (alpha 100).", "Labels: Wake, N1, N2, N3, REM.", "Uniform-guessing reference: 20.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.", "Lightweight balanced quality-screened subset; scores do not describe natural sleep-stage prevalence or the entire73780epoch release.", "Two original corrupt sessions and boundary epochs were excluded by the uploader.", "200Hz data may have finite replacements despite original missing-value flags; source flags are therefore enforced instead of relying on finite checks alone.", "One prediction per30s epoch. Frozen encoders pool15nonoverlapping2s segments; no sequence context across epochs.", "All nights from a participant remain together; pretraining overlap unknown.", "Quality thresholds fixed before any scores; source artifacts may remain." ], "protocolId": "parallel-fixed-subject-folds-v1/eesm19-scalp-sleep", "version": "Hugging Face processed mirror of OpenNeuro ds005185 1.0.2 · {\"mirror\": \"fb045012b58a0b755f9fedf5931dff9b64a9797f\", \"upstream\": \"0857858f7a2ba1582930f23eca3ec56f90a96da9\"}", "elapsed": 114.69470374239609, "peakGb": null, "auditSha": "e6cbf1b199bf2902b9c46e4575ebfa8aff47952625df3d71a1d720b11cc27291", "source": "https://doi.org/10.18112/openneuro.ds005185.v1.0.2", "backend": "Local Ubuntu / CUDA", "chanceLevel": 20.0, "selection": "One fixed seed and training budget; multi-seed sensitivity pending.", "rows": [ { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "channels": 6, "mode": "Supervised fit", "x": 0.7054471302860993, "y": 72.23333333333333, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 70.7–73.8%", "interval": [ 70.7, 73.8 ], "abstain": null, "seconds": 2.584676503902301, "subjects": 20, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." }, { "id": "labram", "name": "LaBraM", "family": "foundation", "parameters": null, "channels": 6, "mode": "Frozen encoder + ridge head", "x": 0.6974569819873635, "y": 71.1, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 69.0–73.2%", "interval": [ 68.96666666666667, 73.23333333333333 ], "abstain": null, "seconds": 11.497591717168689, "subjects": 20, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms" }, { "id": "cbramod", "name": "CBraMod", "family": "foundation", "parameters": null, "channels": 6, "mode": "Frozen encoder + ridge head", "x": 0.7024681739140434, "y": 71.26666666666667, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 68.7–73.7%", "interval": [ 68.7, 73.73333333333333 ], "abstain": null, "seconds": 10.246094243135303, "subjects": 20, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Code: MIT · Weights: Apache-2.0 (official model card)" }, { "id": "eegnet", "name": "EEGNet", "family": "small", "parameters": null, "channels": 6, "mode": "Scratch · 10 epochs", "x": 0.5452404899763517, "y": 56.53333333333334, "xDetail": "Mean across held-out participants", "yDetail": "Descriptive 95% interval: 54.9–58.3%", "interval": [ 54.900000000000006, 58.3 ], "abstain": null, "seconds": 90.3663412781898, "subjects": 20, "note": "One fixed configuration. Foundation encoders remain frozen; small networks train from scratch. These scores do not establish optimal fine-tuned performance. No individual predictions or participant-level results are distributed.", "modelRights": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used." } ], "summarySha": "b7c784c17d6b33c16ffca5df42d3fd8f3d7bf14960675a7278089c4903886567", "protocolSha": "dace8a98e480ae7ee890dd6f498d5753a90e1573813b6e174f3c14bcbae8a8ff", "license": "CC0-1.0 declared by upstream and mirror", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "attribution": "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.", "privacyReview": "Paper reports Central Denmark Region ethics approval 1-10-72-413-17, Danish Medicines Agency approval 2017111085, registration, and written consent. Upstream includes exact age/sex plus sleep diaries, questionnaires, actigraphy, and timing; none should be exposed.", "reviewedAt": "2026-09-20", "reviewBasis": [ "https://doi.org/10.18112/openneuro.ds005185.v1.0.2", "https://www.nature.com/articles/s41598-019-53115-3" ], "rightsScope": "Personal noncommercial research; aggregate results only", "pretrainingOverlap": "Unknown unless explicitly documented; no unseen-pretraining claim." } ], "models": [ { "id": null, "name": "LaBraM", "family": "foundation", "parameters": 5819936, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms", "url": "https://github.com/935963004/LaBraM" }, { "id": null, "name": "EEGPT", "family": "foundation", "parameters": 25287168, "status": "Rights review pending", "note": "Evaluated locally, but no score is published in this release: the checkpoint license is unresolved. Results are withheld pending that review, not because the model failed to run.", "license": "Code: Apache-2.0 · Checkpoint: license unresolved; excluded from new public scope", "url": "https://github.com/BINE022/EEGPT" }, { "id": null, "name": "CBraMod", "family": "foundation", "parameters": 4883800, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Code: MIT · Weights: Apache-2.0 (official model card)", "url": "https://github.com/wjq-learning/CBraMod" }, { "id": null, "name": "EEGNet", "family": "small", "parameters": 1858, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://github.com/braindecode/braindecode/tree/v1.5.1" }, { "id": null, "name": "ShallowFBCSPNet", "family": "small", "parameters": 30482, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://github.com/braindecode/braindecode/tree/v1.5.1" }, { "id": null, "name": "Deep4Net", "family": "small", "parameters": 274552, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://github.com/braindecode/braindecode/tree/v1.5.1" }, { "id": null, "name": "CSP+LDA", "family": "classical", "parameters": null, "status": "Evaluated", "note": "A fixed configuration has been evaluated. See each task for channels and training mode.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://mne.tools/stable/generated/mne.decoding.CSP.html" }, { "id": null, "name": "BIOT", "family": "foundation", "parameters": 3187202, "status": "Adapter needed", "note": "Pretrained bipolar tokens do not directly match the unipolar recordings. A separately validated adapter is required.", "license": "MIT", "url": "https://github.com/ycq091044/BIOT" }, { "id": null, "name": "REVE Base", "family": "foundation", "parameters": null, "status": "Access gated", "note": "Code and electrode positions are accessible. Base weights require accepting the official access agreement; they have not been downloaded.", "license": "MIT", "url": "https://github.com/elouayas/reve_eeg" }, { "id": null, "name": "Signal-JEPA", "family": "foundation", "parameters": 3460000, "status": "Research candidate", "note": "Public, ungated weights; research candidate only; not downloaded or executed for this release. The official model card provides a 13,847,488-byte safetensors checkpoint trained on Lee2019 at 128 Hz. All 17 ds005342 channel names match its 62-channel pretraining layout, but a separate locked adaptation protocol is still required.", "license": "MIT, as stated on the model card", "url": "https://huggingface.co/braindecode/signal-jepa" }, { "id": null, "name": "BENDR", "family": "foundation", "parameters": 157000000, "status": "Research candidate", "note": "Public, ungated weights; research candidate only; not downloaded or executed in this expansion. The Braindecode checkpoint is 628,580,476 bytes and expects 20 channels at 250 Hz. A documented montage or input adapter is required for 17-channel data.", "license": "BSD-3-Clause on the Braindecode model card; verify original release terms before redistributing weights", "url": "https://huggingface.co/braindecode/braindecode-bendr" }, { "id": null, "name": "CSBrain", "family": "foundation", "parameters": null, "status": "Access unverified", "note": "Public source and an official Google Drive weight link are listed; anonymous weight download, file size, and execution remain unverified. The backbone accepts caller-supplied brain regions and channel ordering, while the published BCIC head is fixed to 22 channels. A new adapter and a checkpoint-key loading audit are required.", "license": "Unclear: no LICENSE file was found at the reviewed official repository commit, and no separate weight license was identified", "url": "https://github.com/yuchen2199/CSBrain" }, { "id": null, "name": "Neuro-GPT", "family": "foundation", "parameters": null, "status": "Research candidate", "note": "Public code and Hugging Face weights; research candidate only; not downloaded or executed for this release. The model page shows an approximately 318 MB checkpoint. The published EEGConformer input path is fixed to 22 channels, so using 17 channels requires an explicit mapping or replacement of the spatial layer.", "license": "GPL-3.0 for the official repository and model card", "url": "https://github.com/wenhui0206/NeuroGPT" }, { "id": null, "name": "EEG-Conformer", "family": "small", "parameters": null, "status": "Research candidate", "note": "Public source; no official pretrained checkpoint was found; not executed for this release. This is a from-scratch baseline rather than a foundation-model checkpoint. The author script fixes the spatial convolution to 22 channels and must be parameterized for other montages.", "license": "GPL-3.0", "url": "https://github.com/eeyhsong/EEG-Conformer" }, { "id": null, "name": "EEGSimpleConv", "family": "small", "parameters": null, "status": "Research candidate", "note": "Public implementation; no pretrained weights; not executed for this release. The maintained Braindecode implementation accepts the channel count and sampling rate directly, making it a practical lightweight from-scratch comparator.", "license": "BSD-3-Clause for the Braindecode implementation; the separate author repository license is unclear", "url": "https://braindecode.org/stable/generated/braindecode.models.EEGSimpleConv.html" }, { "id": "cca", "name": "Standard CCA", "family": "classical", "parameters": null, "status": "Evaluated", "note": "Fixed reference method; inspect each task for input features and fit protocol.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://scikit-learn.org/stable/" }, { "id": "spectral-ridge", "name": "Spectral ridge", "family": "classical", "parameters": null, "status": "Evaluated", "note": "Fixed reference method; inspect each task for input features and fit protocol.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://scikit-learn.org/stable/" }, { "id": "temporal-ridge", "name": "Temporal ridge", "family": "classical", "parameters": null, "status": "Evaluated", "note": "Fixed reference method; inspect each task for input features and fit protocol.", "license": "Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.", "url": "https://scikit-learn.org/stable/" } ], "datasets": [ { "name": "ds003810", "task": "Motor imagery / rest", "subjects": 10, "channels": 15, "size": "—", "status": "Aggregate results", "detail": "Clear CC0 snapshot plus dataset-specific ethics and signed-consent evidence; aggregate-only publication materially limits privacy exposure.", "source": "https://doi.org/10.18112/openneuro.ds003810.v2.0.2", "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "evaluated": true, "attribution": "Peterson et al. · OpenNeuro ds003810, version 2.0.2. Study: https://pmc.ncbi.nlm.nih.gov/articles/PMC9114495/", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "EEGMAT", "task": "Arithmetic / rest", "subjects": 36, "channels": 19, "size": "—", "status": "Aggregate results", "detail": "Attribution license and study-specific approval/consent are documented; eligibility assumes aggregate-only output and exclusion of subject-info fields.", "source": "https://physionet.org/content/eegmat/1.0.0/", "license": "Open Data Commons Attribution License 1.0", "licenseUrl": "https://opendatacommons.org/licenses/by/1-0/", "evaluated": true, "attribution": "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.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "EESM19 scalp subset", "task": "Five-stage sleep", "subjects": 20, "channels": 6, "size": "—", "status": "Aggregate results", "detail": "Clear upstream CC0 and study ethics evidence; use only the pinned scalp-signal derivative and publish aggregates, with provenance disclosed.", "source": "https://doi.org/10.18112/openneuro.ds005185.v1.0.2", "license": "CC0-1.0 declared by upstream and mirror", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "evaluated": true, "attribution": "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.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "BETA", "task": "40-target SSVEP", "subjects": 70, "channels": "4 / 8 selected", "size": "—", "status": "Aggregate results", "detail": "Eligible only under the stated personal, research-led, noncommercial operation. Any ads, sponsorship, fees, or use directed toward commercial advantage requires fresh review or permission.", "source": "https://figshare.com/articles/dataset/The_BETA_database/12264401", "license": "CC BY 4.0", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "evaluated": true, "attribution": "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.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "BNCI2014-008", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "Small clinical cohort and linked health attributes create disproportionate reidentification risk; aggregate-only output does not resolve the missing secondary-publication consent evidence.", "source": "https://bnci-horizon-2020.eu/database/data-sets/008-2014", "license": "CC BY-NC-ND 4.0", "licenseUrl": "https://creativecommons.org/licenses/by-nc-nd/4.0/", "evaluated": false, "attribution": "BNCI Horizon 2020 catalog · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "BNCI2014-009", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "License permits only noncommercial unadapted redistribution, and the consent/ethics chain for public secondary results remains unverified.", "source": "https://bnci-horizon-2020.eu/database/data-sets/009-2014", "license": "CC BY-NC-ND 4.0", "licenseUrl": "https://creativecommons.org/licenses/by-nc-nd/4.0/", "evaluated": false, "attribution": "BNCI Horizon 2020 catalog · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "stew-monster-processed", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "Ethics evidence is positive, but upstream license and the processed mirror's authority to apply CC BY remain unresolved.", "source": "https://ieee-dataport.org/open-access/simultaneous-task-eeg-workload-dataset", "license": "Mirror declares CC BY 4.0; exact upstream DataPort license not independently verified", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "evaluated": false, "attribution": "monster-monash/STEW Hugging Face processed mirror · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "TMNRED / ds005383", "task": "Semantic target ERP", "subjects": 30, "channels": 30, "size": "—", "status": "Aggregate results", "detail": "Strong study-specific open-sharing evidence. Aggregate metrics avoid redistribution of stimulus text and participant metadata; credit under CC BY.", "source": "https://doi.org/10.18112/openneuro.ds005383.v1.0.0", "license": "OpenNeuro metadata says CC0; accompanying publication/GitHub says CC BY 4.0", "licenseUrl": "https://creativecommons.org/licenses/by/4.0/", "evaluated": true, "attribution": "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.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "ds006593", "task": "P300 target ERP", "subjects": 21, "channels": 19, "size": "—", "status": "Aggregate results", "detail": "Pinned CC0 release with dataset-specific IRB and consent evidence and comparatively low metadata exposure.", "source": "https://doi.org/10.18112/openneuro.ds006593.v1.0.0", "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "evaluated": true, "attribution": "OpenNeuro ds006593 contributors · version 1.0.0, doi:10.18112/openneuro.ds006593.v1.0.0; original author credits retained at the linked source.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" }, { "name": "physionet-eegmmidb-1.0.0", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "Reuse license is clear, but the dataset-specific human-subject consent and ethics chain is not evidenced strongly enough for a new public benchmark release.", "source": "https://physionet.org/content/eegmmidb/1.0.0/", "license": "Open Data Commons Attribution License 1.0", "licenseUrl": "https://opendatacommons.org/licenses/by/1-0/", "evaluated": false, "attribution": "PhysioNet EEG Motor Movement/Imagery Dataset 1.0.0 · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "BNCI2014-001", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "Per-dataset license is verified, but consent evidence and the application of ND to the planned publication pipeline should be clarified.", "source": "https://bnci-horizon-2020.eu/database/data-sets/001-2014", "license": "CC BY-ND 4.0", "licenseUrl": "https://creativecommons.org/licenses/by-nd/4.0/", "evaluated": false, "attribution": "BNCI Horizon 2020 catalog · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "BNCI2014-004", "task": "Research source", "subjects": "—", "channels": "—", "size": "—", "status": "Not in this release", "detail": "Per-dataset license is verified, but consent evidence and the application of ND to the planned publication pipeline should be clarified.", "source": "https://bnci-horizon-2020.eu/database/data-sets/004-2014", "license": "CC BY-ND 4.0", "licenseUrl": "https://creativecommons.org/licenses/by-nd/4.0/", "evaluated": false, "attribution": "BNCI Horizon 2020 catalog · Original researchers credited at source.", "reviewedAt": "2026-09-20", "decision": "withhold" }, { "name": "ds005342", "task": "Cue-gated idle / command", "subjects": "4 evaluated / 32 prepared", "channels": 17, "size": "—", "status": "Aggregate results", "detail": "Eligible only as an aggregate-only case study: remove all participant IDs/results and condition breakdowns, label n=4 prominently, avoid subgroup claims, and make no claim that the cohort is anonymous.", "source": "https://doi.org/10.18112/openneuro.ds005342.v1.0.3", "license": "CC0-1.0", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "evaluated": true, "attribution": "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.", "reviewedAt": "2026-09-20", "decision": "aggregate_preview" } ], "news": [ { "tag": "Benchmark", "date": "2026-08-03", "title": "A new benchmark compares EEG foundation models with task-specific decoders", "summary": "Researchers surveyed 55 representative models and evaluated 12 open-source foundation models with task-specific baselines across 13 datasets and nine BCI paradigms. The study reports that linear probing is often insufficient and that larger models do not automatically generalize better, reinforcing the need for fixed cross-subject and few-shot calibration protocols.", "source": "Original arXiv paper", "sourceUrl": "https://arxiv.org/abs/2601.17883v3" }, { "tag": "Model release", "date": "2025-10-24", "title": "REVE releases an EEG foundation model for varying electrode layouts", "summary": "REVE reports pretraining on 92 datasets, about 60,000 hours of EEG, and 25,000 participants, with position encodings designed for different electrode arrangements. Its public code, weights, and tutorials make cross-montage transfer an auditable evaluation target. Access to the Base checkpoint is gated; it is not yet available in our local test pool.", "source": "Original arXiv paper and NeurIPS 2025 paper page", "sourceUrl": "https://arxiv.org/abs/2510.21585" }, { "tag": "Research release", "date": "2025-09-18", "title": "NeurIPT brings heterogeneous EEG configurations into a unified pretraining framework", "summary": "NeurIPT models homogeneous and heterogeneous spatiotemporal relationships across BCI tasks and electrode layouts and appeared as a NeurIPS 2025 poster. Its official repository still says that code restoration is in progress, so it is a research update rather than a reproducible model card.", "source": "OpenReview / NeurIPS 2025", "sourceUrl": "https://openreview.net/forum?id=D4hcJPkJ3y" }, { "tag": "Model release", "date": "2025-06-29", "title": "CSBrain adds cross-scale spatiotemporal structure to general EEG decoding", "summary": "CSBrain combines local temporal windows, brain regions, and structured sparse attention, with reported experiments spanning 16 datasets and 11 task types. The official repository provides pretraining and downstream scripts plus a public weight entry point, making it a candidate for a later unified-protocol evaluation.", "source": "Original arXiv paper and NeurIPS 2025 paper page", "sourceUrl": "https://arxiv.org/abs/2506.23075" }, { "tag": "Real-time BCI", "date": "2025-06-02", "title": "An eight-channel, low-preprocessing self-supervised EEG model targets real-time BCI", "summary": "This study applies HuBERT-style self-supervised learning to eight-channel scalp EEG and examines motor imagery, P300, subject differences, and alpha rhythms. It motivates a separate deployment-oriented track for low channel counts, limited preprocessing, and runtime budgets.", "source": "Original arXiv paper", "sourceUrl": "https://arxiv.org/abs/2506.01867" } ], "evidencePolicy": "Only explicitly reviewed cohort-level research results are exported. Raw EEG, individual results and model weights stay outside the website. Each comparison states its source, license, protocol and limitations. No cross-task overall score.", "releaseScope": "Personal noncommercial research · no advertising, paid access or EEG uploads", "coverage": { "displayedComparisons": 39, "displayedProtocols": 8, "completedBatchComparisons": 69, "completedBatchDatasets": 13 } }