1.7%1 / 60 idle trials1 of 4 people always abstained
17
4
13.9 s
TS+LRClassical method · OAS · idle-native windows · Apple M5 / CPU
Supervised fit
35.0%21 / 60 command trials
6.7%4 / 60 idle trials
17
4
—
MDMClassical method · OAS · idle-native windows · Apple M5 / CPU
Supervised fit
35.0%21 / 60 command trials
1.7%1 / 60 idle trials
17
4
—
Scoring time covers fitting and prediction in this configuration, may include accelerator waiting, and excludes data preparation. Scoring time is not recorded for the rows of core-matrix-update-20261106.
Not applicable under this protocol:Standard CCA, FBCCA (they need the stimulation frequencies, and this protocol has none); Temporal ridge, XDAWNCov+TS+LR, xDAWN+LDA (they model the waveform of a stimulus-locked transient potential, and this protocol's classes are not defined by one).
Not run under this protocol:Spectral ridge.
Pick-two protocol comparison
Compare any two methods on Idle & command
Inspect point estimates and 95% intervals side by side on a shared axis. Paired differences appear only where a paired interval is published. Dropdowns are grouped by family alphabetically — never ordered by score.
TS+LR, MDM: Summed test trials. Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.
MDM: MDM's probabilities are a softmax of minus the squared Riemannian distance, an uncalibrated scale, read against the published threshold grid (0.50–0.99).
People who always abstained (TS+LR, MDM): Left empty: counts of people meeting a criterion print only for cohorts of 10 or more.
Idle: differences in summed test trials
Differences in summed test-trial counts against the published idle rows; descriptive, no interval. Against EEGNet: against the matrix's one-seed EEGNet cell; training variance not included. Against CSP+LDA: against the published CSP+LDA idle row (deterministic).
Method
Against EEGNet · Idle trials falsely triggered
Against EEGNet · Command trials detected
Against CSP+LDA · Idle trials falsely triggered
Against CSP+LDA · Command trials detected
TS+LR
+4
+20
+4
+7
MDM
+1
+20
+1
+7
Beside the core matrix
Further foundation encoders, frozen
Each encoder replaces LaBraM or CBraMod in the frozen recipe above; the windows, people, folds, per-person linear head with its calibration-only threshold and scoring stay the same. Rows are grouped by model family; they are counts with no interval, so no row is called above or below another. Every encoder sees this protocol’s published windows, usually shorter than its pretraining context.
10.0%6 of 60 commands2 of 4 people always abstained
8.3%5 of 60 idle trials
50.3%AUROC 0.528
17
4
not in the authors' published pretraining list (checked 2026-10-04)source ↗
ZUNA 1.113Foundation modelModel card: research use only, not for diagnosis or clinical use.
Frozen encoder + linear head
18.3%11 of 60 commands0 of 4 people always abstained
1.7%1 of 60 idle trials
57.2%AUROC 0.608
17
4
Exposure unknownunknown: the authors do not list their pretraining data closely enough to decidesource ↗
Pretraining exposure is what each model’s authors publish as their pretraining list, checked on 2026-10-04: a dataset absent from it is “not in the authors’ published pretraining list”, which is not proof that its recordings were never seen. ST-EEGFormer lists BETA and SingLEM lists TMNRED; ZUNA 1.1 publishes no list, so every one of its cells is unknown; EEGMamba’s list is read from its official code (its paper was not read), with medium confidence.
Row footnotes
Input removes the per-segment per-channel mean before the published /100 scaling (declared before scoring); a no-mean-removal sensitivity run scores P300 0.71 percentage points higher and sleep 1.40 lower.
Input removes the per-segment per-channel mean (declared before scoring); the no-mean-removal sensitivity run scores P300 2.06 percentage points lower and sleep 0.93 lower.
Resampled to 256 Hz and truncated to whole 40-sample patches; 1-2 s windows and 4-8 channel montages are outside the authors' tested regime.
As LUNA Base. Frozen probes only.
p300-target and semantic-target not run: the 2 s tokenizer window cannot be met by 1 s segments without zero padding. Upstream common average reference kept on 4-8 channel montages.
Published CBraMod pooling (patch mean, channels kept); LoRA effectively adapts only the V projection (1-token attention window upstream).
Published CBraMod pooling; LoRA on Mamba2 in_proj and out_proj (24 targets).
BETA cells are pretraining-exposed (BETA is in the ST-EEGFormer corpus).
BETA cells are pretraining-exposed. The authors' downstream class (1-based temporal index) was used; Large may have been pretrained with a 0-based index. Frozen probes only.
Paper-recommended masking geometry; channel-retained pooling (C x 512). Three sibling checkpoints appear in the masking ablation panel.
In design: p300-target and semantic-target. The other six protocols are negative controls outside the model's design. ERP windows have no pre-stimulus baseline.
semantic-target is pretraining-exposed (TMNRED is in the SingLEM corpus). Each channel is encoded independently; the last 0.25 s of each 2 s segment is not seen by the tokenizer.
Pretraining exposure unknown for every core dataset (no published source list). Encoder only; no channel upsampling or reconstruction.
A sibling of the paper-recommended checkpoint in the masking ablation: the same architecture, corpus and recipe, another masking framework or geometry.
Masking ablation: three sibling checkpoints of eeg-fm-masking
Intervals are marginal: the same people under two checkpoints are not tested as a pair.
8.3%5 of 60 commands2 of 4 people always abstained
0.0%0 of 60 idle trials
54.7%AUROC 0.795
17
4
not in the authors' published pretraining list (checked 2026-10-04)source ↗
Weights licences and terms
REVE Base · REVE Large: REVE Responsible Use License v1.0 (owner-approved 2026-10-04)REVE Responsible Use License v1.0: aggregate scientific results only; no re-identification, model inversion, membership inference or non-consensual profiling; cite the REVE paper and name the model version; no adapted weights shared.Model versions: REVE Base @ dc2a075c · REVE Large @ 317531c7.paper ↗
LUNA Base · LUNA Large: CC BY-ND 4.0 (owner-approved: internal adaptation, aggregate scores only, no adapted weights or LoRA deltas shared)CC BY-ND 4.0 weights: scores from internal probing/adaptation only; no modified weights, LoRA or adapter deltas are shared; no implied endorsement by the LUNA authors; the LUNA name is not used to suggest an official release.paper ↗
ST-EEGFormer Base · ST-EEGFormer Large: MIT (repo LICENSE; README reserves the paper, diagrams and the name)MIT (repository LICENSE; the README reserves the paper, diagrams and the ST-EEGFormer name).paper ↗
eeg-fm-masking MAE (r=9 cm, L=2) · eeg-fm-masking JEPA (r=9 cm, L=2) · eeg-fm-masking MAE (r=one channel, L=1) · eeg-fm-masking JEPA (r=one channel, L=1): CC-BY-4.0 (cite arXiv:2609.33487)CC-BY-4.0 weights: attribution required (cite arXiv:2609.33487).paper ↗
SingLEM: MIT (checkpoint committed in the MIT repository)MIT.paper ↗
ZUNA 1.1: Apache-2.0; model card: research use only, not for diagnosis or clinical useApache-2.0 weights; model card states research use only, not for diagnosis or clinical use.paper ↗
No model’s authors endorse these results, and no adapted weights or LoRA deltas were kept or shared.
Notes on these rows
REVE Base: REVE input removes the per-segment per-channel mean (declared before scoring; the published LaBraM/CBraMod rows did not); a no-mean-removal sensitivity run is stated in the row footnote.
REVE Large: REVE input removes the per-segment per-channel mean (declared before scoring); a no-mean-removal sensitivity run is stated in the row footnote.
LUNA Base: LUNA resamples to 256 Hz and keeps whole 40-sample patches: 1 s keeps 240/256 samples, 2 s keeps 480/512.
LUNA Large: LUNA resamples to 256 Hz and keeps whole 40-sample patches: 1 s keeps 240/256 samples, 2 s keeps 480/512.
ST-EEGFormer Large: ST-EEGFormer Large uses the authors' downstream class (1-based temporal index); a label-free reconstruction check suggests Large was pretrained with a 0-based index, so its temporal sinusoid may be shifted by one position.
ERP-FM Base: Outside the ERP-FM design: negative control.
SingLEM: SingLEM tokenizer (128-sample tokens, stride 96 at 128 Hz) leaves the last 0.25 s of every 2 s segment unseen.
ZUNA 1.1: ZUNA pretraining sources are not published: exposure unknown on every core dataset. 0-4 electrode coordinates per protocol are clamped to the edge of the +-0.12 m position grid as upstream does.
EEGPT ran in the same run as the published LaBraM row above, with only the encoder swapped: the same windows, people, folds, per-person logistic head with its calibration-only rule and scoring. Its figures are counts and cohort means with no interval, so it is not called above or below another row.
11.7%7 of 60 commands1 of 4 people always abstained
3.3%2 of 60 idle trials
57.2%AUROC 0.780
17
4
not in the authors' published pretraining list (checked 2026-10-12)source ↗
Footnotes to this row
Same run as the published LaBraM idle row; the published CBraMod idle row comes from another run, so a comparison with it is across runs. The window balanced accuracy and AUROC were computed from this run’s saved probabilities with the definitions of the further foundation-encoder rows; the run itself recorded the counts.
Every core window is shorter than EEGPT’s 4-second pretraining window: 1 or 2 seconds, and sleep as fifteen 2-second segments. The encoder computes each 0.25-second patch the same way at any length; a shorter window gives fewer patches to average.
The readout is the site’s fixed probe on the frozen summary tokens, as for the published LaBraM and CBraMod rows: ridge (alpha 100) on 2,048 pooled values, and on idle a per-person logistic head. It is not the authors’ downstream probe, a learnable spatial filter with a linear head, and the windows are fed without the average reference EEGPT was pretrained with.
EEGPT’s exposure is read from its authors’ published pretraining list, checked on 2026-10-12, as for the further encoders. Its pretraining corpus: Five public datasets, as the paper’s Table 1 lists them: PhysioMI (PhysioNet motor imagery and execution), HGD, TSU (the Tsinghua 40-target SSVEP benchmark), SEED and M3CV; 58 electrodes, 256 Hz, millivolts, average reference, 4-second windows.
Licence and terms
Code licence: Apache-2.0 (the repository’s source code only)
Weights licence: unresolved: the checkpoint’s own licence and terms could not be read from its hosting record, and the repository’s code licence is not evidence that it covers the weights
Weights: no original or converted weights are distributed; the hashes identify the checkpoint used and grant nothing
Rights review: Published by the owner’s decision, aggregate results only, as the pilot comparison is. The checkpoint’s licence could not be read; no publisher prohibition on independently computed aggregate results was found; no original or converted weights are distributed.
Six LaBraM at Scale encoders and REVE Small’s encoder in the same frozen recipe as the rows above: the same windows, people, folds, head and scoring, with no random-initialization control.
Cohort size n = 4 (a protocol fact): idle trials with a false trigger, command trials detected within 3 s, and the window metrics.
Encoder
Idle false triggers
Commands detected ≤3 s
Mean window balanced accuracy
Mean window AUROC
LaBraM at Scale · Quark ladder
Quark-BLocal Ubuntu / CPU
4 / 60
15 / 60
60.8%
0.724
Quark-LLocal Ubuntu / CPU
2 / 60
13 / 60
61.1%
0.655
Quark-HLocal Ubuntu / CPU
2 / 60
15 / 60
62.8%
0.746
Quark-GLocal Ubuntu / CUDA
2 / 60
10 / 60
57.2%
0.722
LaBraM at Scale · Atom-L and Molecule-L
Atom-LLocal Ubuntu / CPU
2 / 60its pretraining corpus includes this protocol's dataset
20 / 60
66.1%
0.721
Molecule-LLocal Ubuntu / CPU
1 / 60its pretraining corpus includes this protocol's dataset
10 / 60
60.6%
0.723
REVE Small · early checkpoint (epoch 15 of 200)
REVE SmallLocal Ubuntu / CPU
1 / 60
8 / 60
54.4%
0.680
Idle is a within-person protocol: each row gives the cohort size, trial counts and window metrics, with no contrast. LaBraM at Scale's pooled features are global on idle; the published LaBraM idle row kept the channel axis.
Unless a cell is flagged, its dataset is not in that collection’s published pretraining list (the LaBraM at Scale pretraining manifest at faa387d9, checked 2026-11-05); REVE Small’s is read from the REVE corpus list (checked 2026-10-04).
Terms: LaBraM at Scale’s code and weights under MIT (Pan-Junyu/LaBraM-at-Scale @ a0b0ce27); REVE Small under the REVE Responsible Use License v1.0 (reve-small @ 3565bdcb), aggregate scientific results only. No weights are distributed.
Detection resets at each cue. Only three minutes of test idle are observed; these results cannot estimate natural, always-on false activations per hour.
The protocol, step by step
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.
Stability
Single fixed configuration; no test-set hyperparameter selection.
Pretraining exposure
Unknown unless explicitly documented; no unseen-pretraining claim.
Fixed calibration rule; 60 training trials and 30 calibration trials per participant. Four-person case study; cohort aggregates only, no subgroup or population claim.
Model terms
CSP+LDA: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
LaBraM: Code/repository: MIT · Checkpoint: committed in that repository; no separate weight terms
EEGNet: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
ShallowFBCSPNet: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
CBraMod: Code: MIT · Weights: Apache-2.0 (official model card)
Deep4Net: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
TS+LR: Deterministic; no weights. pyRiemann 0.12 and scikit-learn, BSD-3-Clause.
MDM: Deterministic; no weights. pyRiemann 0.12 and scikit-learn, BSD-3-Clause.
Source and licence
ds005342
CreditOpenNeuro 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.
Public-data register noteEligible 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.