ShallowFBCSPNetCompact model · matrix recipe · AdamW · Local Ubuntu / CUDA
Scratch · 10 epochs
55.0%53.1%–56.8%
0.519Mean across held-out participants
6
20
—
Deep4NetCompact model · matrix recipe · AdamW · Local Ubuntu / CUDA
Scratch · 10 epochs
45.2%42.7%–47.7%
0.419Mean across held-out participants
6
20
—
CSP+LDAClassical method · one-vs-rest CSP · shrinkage LDA · Local Ubuntu / CPU
Supervised fit
50.1%46.9%–53.1%
0.483Mean across held-out participants
6
20
—
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).
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.
CSP+LDA: One-vs-rest CSP: one binary CSP per stage, 30 log-variance features, one shrinkage LDA; broadband as in the published mi-rest recipe; a new code path, not a tuned CSP.
Paired differences
Each difference is the mean over people of each person’s difference in balanced accuracy, in percentage points (pp), on the same windows and people.
ShallowFBCSPNet, Deep4Net and CSP+LDA under the matrix recipe
against the matrix's one-seed EEGNet cell; training variance not included
Difference
Mean (pp)
95% interval, unadjusted
Bonferroni interval over 18 contrasts
Reading
CSP+LDA − EEGNetCSP+LDA (one-vs-rest CSP)
−6.4 pp
−9.3 to −3.3
−10.8 to −1.5
lower
Deep4Net − EEGNet
−11.3 pp
−13.6 to −9.0
−14.7 to −7.8
lower
ShallowFBCSPNet − EEGNet
−1.6 pp
−3.4 to 0.4
−4.7 to 1.5
unresolved
The paired 95 % interval governs the wording; the Bonferroni interval beside it covers all 18 contrasts. Read at 95 % without correction, about one in 20 contrasts with no true difference may show one by chance.
Re-training on the CUDA card is not bit-exact; re-training the published semantic-target EEGNet cell there moved it by 0.1 pp.
Beside the core matrix
Log band-power ridge (5 bands, visible channels)
Five log band powers (1–4, 4–8, 8–13, 13–30 and 30–45 Hz) per channel on all recorded channels of the protocol, from one Hann periodogram of each published window after removing each channel's window mean and a per-window robust amplitude scaling, read by the core ridge head (alpha 100) on the published windows and participant-held-out folds. It uses the same windows, people, folds and ridge head as the Spectral ridge row, repeated from the table above.
Each difference is this row minus the named row on the same people, with a paired participant bootstrap (10,000 draws): an interval wholly above zero reads above, wholly below reads below, and one that includes zero reads not separated, which does not show the two are equivalent. The intervals are descriptive, conditional on these folds, with no multiplicity correction.
Minus the Spectral ridge row: −6.23 pp (−8.50 to −3.90): below. The two rows differ in bands, normalisation, estimator, frequency range and sampling rate at once, so the difference measures the whole recipe.
Minus ST-EEGFormer Large, the highest of the 14 frozen-encoder rows this comparison was specified against, chosen on these same test folds: −13.60 pp (−15.80 to −11.37): below. Choosing it on the test folds leans the comparison against this row.
Each encoder replaces LaBraM or CBraMod in the frozen recipe above; the windows, people, folds, ridge head and scoring stay the same. Rows are grouped by model family; a row is above or below another only where their 95% intervals do not overlap. Each interval is a descriptive participant bootstrap with no multiplicity correction, and with 126 frozen cells an occasional non-overlap is expected by chance. Every encoder sees this protocol’s published windows, usually shorter than its pretraining context.
Spectral ridge
72.2% (70.7%–73.8%)
LaBraM
71.1% (69.0%–73.2%)
CBraMod
71.3% (68.7%–73.7%)
EEGNet
56.5% (54.9%–58.3%)
REVE Base
74.6% (72.9%–76.4%)
REVE Large
76.6% (75.2%–78.0%)
LUNA Base
67.5% (65.3%–69.5%)
LUNA Large
72.4% (70.4%–74.0%)
BrainOmni Base
67.7% (65.8%–69.5%)
CodeBrain
70.7% (68.8%–72.7%)
EEGMamba
64.1% (61.6%–66.4%)
ST-EEGFormer Base
77.3% (75.6%–78.8%)
ST-EEGFormer Large
79.6% (78.2%–81.0%)
eeg-fm-masking MAE (r=9 cm, L=2)
67.9% (65.4%–70.4%)
ERP-FM Base
71.0% (69.5%–72.6%)
SingLEM
35.3% (33.7%–37.0%)
ZUNA 1.1
60.1% (57.8%–62.3%)
0%25%50%75%100%
Balanced accuracy under this protocol. Grey: the released rows above; rust: the further encoders, in family order. Dot: the estimate; line: 95% interval; dashed line: chance level.
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 + ridge head
60.1%57.8%–62.3%
0.579
6
20
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.
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 min-max position scaling on a 6-channel montage.
LUNA Large: LUNA resamples to 256 Hz and keeps whole 40-sample patches: 1 s keeps 240/256 samples, 2 s keeps 480/512. LUNA min-max position scaling on a 6-channel montage.
BrainOmni Base: BrainOmni common average reference on 6 channels removes a median 0.45 of segment channel variance.
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 runs as the published LaBraM and CBraMod rows above, with only the encoder swapped: the same windows, people, folds, ridge head and scoring.
Spectral ridge
72.2% (70.7%–73.8%)
LaBraM
71.1% (69.0%–73.2%)
CBraMod
71.3% (68.7%–73.7%)
EEGNet
56.5% (54.9%–58.3%)
EEGPT
75.5% (73.9%–77.1%)
0%25%50%75%100%
Balanced accuracy under this protocol. Grey: the released rows above; rust: EEGPT. Dot: the estimate; line: 95% interval; dashed line: chance level.
not in the authors' published pretraining list (checked 2026-10-12)source ↗
Footnotes to this row
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.
Against the published rows
Each row of the table above that experiments.json holds, by whether its 95% interval and EEGPT’s overlap (no multiplicity correction):
Spectral ridge: EEGPT above, the intervals 0.10 pp apart
LaBraM: EEGPT above, the intervals 0.67 pp apart
CBraMod: EEGPT above, the intervals 0.17 pp apart
EEGNet: EEGPT above, the intervals 15.60 pp apart
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.
Balanced accuracy with its 95% interval, and macro F1, frozen encoders.
† Quark-G: Run on Local Ubuntu / CUDA; its step is read against a GPU re-run of Quark-H and includes the probe's input width doubling from 800 to 1600 under one fixed penalty (see the topic page).
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.
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.
The protocol, step by step
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).
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.
Stability
One fixed seed and training budget; multi-seed sensitivity pending.
Pretraining exposure
Unknown unless explicitly documented; no unseen-pretraining claim.
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.
Model terms
Spectral ridge: 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
CBraMod: Code: MIT · Weights: Apache-2.0 (official model card)
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.
Deep4Net: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
CSP+LDA: Trained from scratch / deterministic reference; no third-party pretrained weights. Braindecode BSD-3-Clause; MNE/scikit-learn BSD where used.
Source and licence
EESM19 scalp subset
CreditKaare 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.
The figures here are aggregate measurements computed by BCI Report under the licence above; the data remain their creators’.
Permitted scopePersonal noncommercial research; aggregate results only
Privacy Only cohort aggregates are published here: no recording, no participant identifier, no per-person score. According to the 2025 data descriptor, the informed consent form did not mention publication, and before release the GDPR office of Region Midt judged the data fully anonymised: consent covered the study, and the public release rests on that anonymisation judgement.The full review note is in the protocol JSON ↓
Public-data register noteClear upstream CC0 and study ethics evidence; use only the pinned scalp-signal derivative and publish aggregates, with provenance disclosed. Consent covered the study, not publication; the public release rests on a GDPR anonymization assessment (amended 2026-09-22).
One training run per cell with the matrix seed and the matrix recipe: AdamW, 20 epochs (sleep 10), no schedule, no early stopping. This is the recipe of the EEGNet cells; it is not tuned for any network and it is not the original papers' training. Intervals describe variation across people, not across training runs.