Core-matrix protocol · Transfer to a new person

Semantic target ERP on TMNRED / ds005383

The released protocol, with the score of every method run under it.

The protocol at a glance

Task
Reading · target versus nontarget events
Evaluation
Transfer to a new person
People
30
Data
3,600 epochs · 5 participant-disjoint folds
Input
30 channels · 1-second windows
Chance level
50.0%
Metrics
Balanced accuracy (primary) · Macro F1 (secondary)

Results

Compare within this protocol only: other protocols differ in cohort, electrodes, window or chance level.

Spectral ridge
51.0% (49.6%–52.5%)
Temporal ridge
56.3% (54.4%–58.4%)
LaBraM
53.2% (51.3%–54.9%)
CBraMod
55.8% (54.2%–57.4%)
EEGNet
61.4% (58.9%–64.0%)
XDAWNCov+TS+LR
60.3% (58.2%–62.6%)
xDAWN+LDA
56.8% (55.1%–58.6%)
ShallowFBCSPNet
56.7% (54.7%–58.8%)
Deep4Net
57.4% (55.4%–59.5%)
CSP+LDA
52.7% (51.4%–54.0%)
Balanced accuracy, every method under this protocol. Dot: the estimate; line: descriptive 95% interval; dashed line: chance level.
MethodTraining modeBalanced accuracyMacro F1ChannelsPeopleScoring time
Spectral ridgeClassical methodSupervised fit51.0%49.6%–52.5%Interval reaches chance level0.506Mean across held-out participants30300.9 s
Temporal ridgeClassical methodSupervised fit56.3%54.4%–58.4%0.559Mean across held-out participants30301.2 s
LaBraMFoundation modelFrozen encoder + ridge head53.2%51.3%–54.9%0.519Mean across held-out participants30303.4 s
CBraModFoundation modelFrozen encoder + ridge head55.8%54.2%–57.4%0.554Mean across held-out participants303023.5 s
EEGNetCompact modelScratch · 20 epochs61.4%58.9%–64.0%0.606Mean across held-out participants303083.4 s
XDAWNCov+TS+LRClassical method · 4 filters · 50 Hz · primary row · Local Ubuntu / CPUSupervised fit60.3%58.2%–62.6%0.589Mean across held-out participants3030—
xDAWN+LDAClassical method · 4 filters · 50 Hz · secondary row · Local Ubuntu / CPUSupervised fit56.8%55.1%–58.6%0.562Mean across held-out participants3030—
ShallowFBCSPNetCompact model · matrix recipe · AdamW · Local Ubuntu / CUDAScratch · 20 epochs56.7%54.7%–58.8%0.553Mean across held-out participants3030—
Deep4NetCompact model · matrix recipe · AdamW · Local Ubuntu / CUDAScratch · 20 epochs57.4%55.4%–59.5%0.559Mean across held-out participants3030—
CSP+LDAClassical method · broadband CSP · shrinkage LDA · Local Ubuntu / CPUSupervised fit52.7%51.4%–54.0%0.493Mean across held-out participants3030—

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).

Not run under this protocol: TS+LR, MDM.

Pick-two protocol comparison

Compare any two methods on Semantic target ERP

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.

0%25%50%75%100%Chance 50.0%XDAWNCov+TS+LR60.3% [58.2–62.6]Temporal ridge56.3% [54.4–58.4]
Published paired contrastXDAWNCov+TS+LR − Temporal ridge: +3.9 pp[+2.0, +5.9 pp]

nominal 95% interval · n = 30 participants · Paired interval excludes zero

Default view (Semantic target ERP): XDAWNCov+TS+LR balanced accuracy 60.3% (95% interval 58.2–62.6%); Temporal ridge 56.3% (95% interval 54.4–58.4%); chance 50.0%.

Published paired contrasts table (17 pairs)
Pair (left minus right)Mean differenceGoverning intervalParticipantsAdjustment & notes
XDAWNCov+TS+LR − Spectral ridge+9.3 pp[+6.1, +12.8 pp]3098.75% interval, Bonferroni-adjusted for four contrasts
XDAWNCov+TS+LR − EEGNet−1.2 pp[−3.0, +0.6 pp]30nominal 95% intervalagainst the matrix's one-seed EEGNet cell; training variance not included
xDAWN+LDA − EEGNet−4.6 pp[−6.5, −2.7 pp]30nominal 95% intervalagainst the matrix's one-seed EEGNet cell; training variance not included
xDAWN+LDA − Spectral ridge+5.8 pp[+4.0, +7.8 pp]30nominal 95% interval
XDAWNCov+TS+LR − Temporal ridge+3.9 pp[+2.0, +5.9 pp]30nominal 95% interval
xDAWN+LDA − Temporal ridge+0.5 pp[−1.2, +2.1 pp]30nominal 95% interval
XDAWNCov+TS+LR − LaBraM+7.1 pp[+4.8, +9.4 pp]30nominal 95% interval
XDAWNCov+TS+LR − CBraMod+4.5 pp[+2.5, +6.4 pp]30nominal 95% interval
XDAWNCov+TS+LR − xDAWN+LDA+3.5 pp[+1.8, +5.3 pp]30nominal 95% interval
CSP+LDA − EEGNet−8.8 pp[−11.1, −6.4 pp]3095% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
Deep4Net − EEGNet−4.0 pp[−5.8, −2.1 pp]3095% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
ShallowFBCSPNet − EEGNet−4.7 pp[−7.2, −2.2 pp]3095% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
ShallowFBCSPNet − Deep4Net−0.7 pp[−3.0, +1.5 pp]3095% interval
Log band-power ridge − Spectral ridge+5.1 pp[+3.1, +7.1 pp]3095% paired participant bootstrap interval, unadjusted
LaBraM at Scale Quark-L − LaBraM at Scale Quark-B−0.6 pp[−2.8, +1.7 pp]3095% paired participant bootstrap interval, unadjusted (21 ladder contrasts)
LaBraM at Scale Quark-H − LaBraM at Scale Quark-L−2.6 pp[−5.1, −0.1 pp]3095% paired participant bootstrap interval, unadjusted (21 ladder contrasts)
LaBraM at Scale Quark-H − LaBraM at Scale Quark-B−3.3 pp[−5.5, −0.9 pp]3095% paired participant bootstrap interval, unadjusted (21 ladder contrasts)

How these rows were run

Riemannian and xDAWN baselines

Recipe
XDAWNCov+TS+LR: 4 filters · 50 Hz · primary row; xDAWN+LDA: 4 filters · 50 Hz · secondary row
Hardware
XDAWNCov+TS+LR, xDAWN+LDA: Local Ubuntu / CPU

One fixed configuration per method: library or MOABB-reference settings with the stated deviations. The 4–40 Hz band-pass (or demeaning plus 50 Hz decimation on the ERP protocols) is part of the recipe and is applied to each window on its own.

ShallowFBCSPNet, Deep4Net and CSP+LDA under the matrix recipe

Recipe
ShallowFBCSPNet, Deep4Net: matrix recipe · AdamW; CSP+LDA: broadband CSP · shrinkage LDA
Hardware
ShallowFBCSPNet, Deep4Net: Local Ubuntu / CUDA; CSP+LDA: Local Ubuntu / CPU

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.

Riemannian and xDAWN baselines

DifferenceMean (pp)Interval95% interval, unadjustedReading
XDAWNCov+TS+LR − Spectral ridgeXDAWNCov+TS+LR · 4 filters · 50 HzSpectral ridge: 51.0%+9.3 pp6.1 to 12.898.75% interval, Bonferroni-adjusted for four contrasts6.7 to 12.0higher
XDAWNCov+TS+LR − Temporal ridgeXDAWNCov+TS+LR · 4 filters · 50 HzTemporal ridge: 56.3%+3.9 pp2.0 to 5.9nominal 95% interval—higher

Paired participant bootstrap (10,000 draws). The four contrasts against the Spectral ridge cells carry 98.75 % intervals (Bonferroni over four); the other contrasts carry nominal 95 % intervals with no familywise claim. 'Higher' or 'lower' means the interval lies wholly above or below zero; 'unresolved' means it includes zero.

ShallowFBCSPNet, Deep4Net and CSP+LDA under the matrix recipe

against the matrix's one-seed EEGNet cell; training variance not included
DifferenceMean (pp)95% interval, unadjustedBonferroni interval over 18 contrastsReading
CSP+LDA − EEGNet−8.8 pp−11.1 to −6.4−12.2 to −5.2lower
Deep4Net − EEGNet−4.0 pp−5.8 to −2.1−6.8 to −1.1lower
ShallowFBCSPNet − EEGNet−4.7 pp−7.2 to −2.2−8.5 to −1.1lower

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.

MethodTraining modeBalanced accuracyMacro F1ChannelsPeople
Spectral ridgeClassical method · core matrix row aboveSupervised fit51.03%49.56%–52.53%Interval reaches chance level0.5063030
Log band-power ridge (5 bands, visible channels)Classical methodSupervised fit56.11%54.22%–58.00%0.5463030

Differences from two published rows

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: +5.08 pp (+3.06 to +7.06): above. The two rows differ in bands, normalisation, estimator, frequency range and sampling rate at once, so the difference measures the whole recipe.

Minus REVE Large, the highest of the 13 frozen-encoder rows this comparison was specified against, chosen on these same test folds: −3.56 pp (−5.61 to −1.53): below. Choosing it on the test folds leans the comparison against this row.

Notes on this comparator

On these 1-second windows the 1–4 Hz band partly carries slow drift within the window and the slow ERP itself, as the Spectral ridge row’s lowest bins do.

Beside the core matrix

Further foundation encoders, frozen

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
51.0% (49.6%–52.5%)
Temporal ridge
56.3% (54.4%–58.4%)
LaBraM
53.2% (51.3%–54.9%)
CBraMod
55.8% (54.2%–57.4%)
EEGNet
61.4% (58.9%–64.0%)
REVE Base
53.7% (51.8%–55.4%)
REVE Large
59.7% (57.5%–61.8%)
LUNA Base
56.2% (54.2%–58.2%)
LUNA Large
57.1% (55.3%–58.9%)
CodeBrain
54.7% (53.0%–56.6%)
EEGMamba
54.4% (52.5%–56.3%)
ST-EEGFormer Base
54.6% (52.9%–56.4%)
ST-EEGFormer Large
57.8% (56.0%–59.6%)
eeg-fm-masking MAE (r=9 cm, L=2)
54.8% (52.7%–56.9%)
ERP-FM Base
53.3% (51.7%–54.8%)
SingLEM
58.5% (56.1%–61.1%)
ZUNA 1.1
53.9% (52.2%–55.6%)
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.
EncoderTraining modeBalanced accuracyMacro F1ChannelsPeoplePretraining exposure
REVE Base1Foundation modelFrozen encoder + ridge head53.7%51.8%–55.4%0.5313030not in the authors' published pretraining list (checked 2026-10-04)source ↗
REVE Large2Foundation modelFrozen encoder + ridge head59.7%57.5%–61.8%0.5863030not in the authors' published pretraining list (checked 2026-10-04)source ↗
LUNA Base3Foundation modelFrozen encoder + ridge head56.2%54.2%–58.2%0.5533030not in the authors' published pretraining list (checked 2026-10-04)source ↗
LUNA Large4Foundation modelFrozen encoder + ridge head57.1%55.3%–58.9%0.5663030not in the authors' published pretraining list (checked 2026-10-04)source ↗
BrainOmni Base5Foundation modelFrozen encoder + ridge headNot runBrainTokenizer needs a 512-sample window (2.0 s at 256 Hz); the published segment is 1 s = 256 samples, so upstream unfold() would zero-pad 256 of 512 samples of every window (invented samples). All upstream downstream tasks use >= 2 s windows.30—not in the authors' published pretraining list (checked 2026-10-04)source ↗
CodeBrain6Foundation modelFrozen encoder + ridge head54.7%53.0%–56.6%0.5413030not in the authors' published pretraining list (checked 2026-10-04)source ↗
EEGMamba7Foundation modelFrozen encoder + ridge head54.4%52.5%–56.3%0.5393030not in the authors' published pretraining list (checked 2026-10-04)source ↗
ST-EEGFormer Base8Foundation modelFrozen encoder + ridge head54.6%52.9%–56.4%0.5403030not in the authors' published pretraining list (checked 2026-10-04)source ↗
ST-EEGFormer Large9Foundation modelFrozen encoder + ridge head57.8%56.0%–59.6%0.5613030not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking MAE (r=9 cm, L=2)10Foundation modelFrozen encoder + ridge head54.8%52.7%–56.9%0.5423030not in the authors' published pretraining list (checked 2026-10-04)source ↗
ERP-FM Base11Foundation modelFrozen encoder + ridge head53.3%51.7%–54.8%0.5273030not in the authors' published pretraining list (checked 2026-10-04)source ↗
SingLEM12Foundation modelFrozen encoder + ridge head58.5%56.1%–61.1%0.5783030in the authors' published pretraining listsource ↗
ZUNA 1.113Foundation modelModel card: research use only, not for diagnosis or clinical use.Frozen encoder + ridge head53.9%52.2%–55.6%0.5343030Exposure 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

  1. 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.
  2. 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.
  3. 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.
  4. As LUNA Base. Frozen probes only.
  5. 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.
  6. Published CBraMod pooling (patch mean, channels kept); LoRA effectively adapts only the V projection (1-token attention window upstream).
  7. Published CBraMod pooling; LoRA on Mamba2 in_proj and out_proj (24 targets).
  8. BETA cells are pretraining-exposed (BETA is in the ST-EEGFormer corpus).
  9. 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.
  10. Paper-recommended masking geometry; channel-retained pooling (C x 512). Three sibling checkpoints appear in the masking ablation panel.
  11. 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.
  12. 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.
  13. Pretraining exposure unknown for every core dataset (no published source list). Encoder only; no channel upsampling or reconstruction.
  14. 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.

EncoderTraining modeBalanced accuracyMacro F1ChannelsPeoplePretraining exposure
eeg-fm-masking MAE (r=9 cm, L=2)10Foundation model · matrix row aboveFrozen encoder + ridge head54.8%52.7%–56.9%0.5423030not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking JEPA (r=9 cm, L=2)14Foundation modelFrozen encoder + ridge head55.8%53.9%–57.6%0.5493030not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking MAE (r=one channel, L=1)14Foundation modelFrozen encoder + ridge head54.9%53.4%–56.4%0.5423030not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking JEPA (r=one channel, L=1)14Foundation modelFrozen encoder + ridge head55.0%53.4%–56.6%0.5433030not in the authors' published pretraining list (checked 2026-10-04)source ↗

Weights licences and terms

No model’s authors endorse these results, and no adapted weights or LoRA deltas were kept or shared.

Notes on these rows

Beside the core matrix

EEGPT, frozen

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
51.0% (49.6%–52.5%)
Temporal ridge
56.3% (54.4%–58.4%)
LaBraM
53.2% (51.3%–54.9%)
CBraMod
55.8% (54.2%–57.4%)
EEGNet
61.4% (58.9%–64.0%)
EEGPT
54.6% (53.3%–55.9%)
Balanced accuracy under this protocol. Grey: the released rows above; rust: EEGPT. Dot: the estimate; line: 95% interval; dashed line: chance level.
EncoderTraining modeBalanced accuracyMacro F1ChannelsPeoplePretraining exposure
EEGPTFoundation modelFrozen encoder + ridge head54.6%53.3%–55.9%0.5403030not in the authors' published pretraining list (checked 2026-10-12)source ↗

Footnotes to this row

  1. On the P300 and semantic targets, one-second windows ran as four real 0.25-second patches: the patch grid was set to four after the encoder was built, with no padding. EEGPT’s encoder treats each time patch on its own, so this is numerically the same as building it for 256-sample inputs, as the Orthosis pilot built it for its 128-sample windows.
  2. 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.
  3. 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):

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

Frozen-probe comparison

LaBraM at Scale and REVE Small, frozen

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.
EncoderBalanced accuracy (95% interval)Macro F1Exposure
LaBraM at Scale · Quark ladder
Quark-BLocal Ubuntu / CPU55.4%53.1%–57.8%0.543—
Quark-LLocal Ubuntu / CPU54.8%53.0%–56.6%0.543—
Quark-HLocal Ubuntu / CPU52.2%50.3%–53.9%0.517—
Quark-G†Local Ubuntu / CUDA51.8%50.1%–53.4%0.514—
LaBraM at Scale · Atom-L and Molecule-L
Atom-LLocal Ubuntu / CPU52.8%51.3%–54.3%0.517its pretraining corpus includes this protocol's dataset
Molecule-LLocal Ubuntu / CPU54.7%52.9%–56.5%0.542its pretraining corpus includes this protocol's dataset
REVE Small · early checkpoint (epoch 15 of 200)
REVE SmallLocal Ubuntu / CPU56.7%54.5%–59.0%0.558—

† 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).

The Quark ladder’s paired differences across protocols, the Quark-G step among them, are on “Does pretraining help EEG foundation models like LaBraM and CBraMod?”.

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.

Limits

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.

The protocol, step by step

  1. 5 participant-disjoint folds. All recordings from a person stay together. Each person contributes to the held-out predictions once.
  2. units: microvolts; epoch: [event onset, onset + 1.0 s); filtering: none added
  3. 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).
  4. Labels: nontarget, target.
  5. Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.
  6. Research subset with fixed deterministic sampling capped at 60 epochs per participant/class.
  7. Balancing to 60 per class changes the source target/nontarget prevalence.
  8. Events use numbered target/nontarget variants; the benchmark collapses the suffix only because the trial_type prefix explicitly names the semantic class.

Stability

One fixed seed and training budget; multi-seed sensitivity pending.

Pretraining exposure

Unknown unless explicitly documented; no unseen-pretraining claim.

Checked 2026-10-04. LaBraM: not in the authors' published pretraining list (checked 2026-10-04) source ↗; CBraMod: not in the authors' published pretraining list (checked 2026-10-04) source ↗. The released sentence above predates this check. Further encoders: the exposure column of their table above → · EEGPT: its own section above →

Notes on the methods

Model terms

Source and licence

Permitted scope Personal noncommercial research; aggregate results only

Privacy Only cohort aggregates are published here: no recording, no participant identifier, no per-person score. The full review note is in the protocol JSON ↓

Public-data register note Strong study-specific open-sharing evidence. Aggregate metrics avoid redistribution of stimulus text and participant metadata; credit under CC BY.

Rights reviewed · against doi.org ↗ · www.nature.com ↗

Reproducibility record

Protocol id
parallel-fixed-subject-folds-v1/ds005383
Dataset release
OpenNeuro snapshot 1.0.0 · ad78f3db430e636595b1b1c08417492f3067a25e
Hardware
Local Ubuntu / CUDA
Scoring stage sum
112.4 s · may include accelerator waiting
Audit record SHA-256
5920974ae9ca3876ce1be82497c139f0783d2ff47173641e329af976441de4c4
Summary SHA-256
c5dca7fc30d07a001b8980a4ea65e55ef292f0891a4d32a12262b8ec88389243
Protocol SHA-256
839dfba89cf200aac7a28a7259f45024d006fcd4500170e247fcc1c4e9962c8d

Rows of core-matrix-update-20261106

Riemannian and xDAWN baselines

Protocol SHA-256
4310329a2d6dddaeff64a965750b64237b08220fc23717b1f3d0c1a48f56b055
Audit record SHA-256
ce1ee2a0345358c613aa89298c1cb90939df15d68d6c48298b99f9daa7f69e82
Hardware
Local Ubuntu / CPU
Row intervals
participant bootstrap, 10,000 draws
Paired-difference intervals
paired participant bootstrap, 10,000 draws

ShallowFBCSPNet, Deep4Net and CSP+LDA under the matrix recipe

Protocol SHA-256
a216bfac4ce6dd11c63d499b58bc93aebb89025168911b5d7d7fe265a866da4c
Audit record SHA-256
f5a1c3dab4a7e265fa3413ada48edc89b9fd4ac98ab7dc0d0f9279761e7e0a9e
Hardware
Local Ubuntu / CUDA · Local Ubuntu / CPU
Row intervals
participant bootstrap, 10,000 draws
Paired-difference intervals
paired participant bootstrap, 10,000 draws

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.

Downloads

This protocol's rows in core-matrix-update-20261106 · CSV ↓ · Core matrix: rows run under its exact protocols · JSON ↓

Other protocols

All protocols → · The core matrix on the home page → · All datasets → · All methods →

Cite this page

BCI Report (2026). Semantic target ERP on TMNRED / ds005383. https://bci.report/protocols/semantic-target/

Figures from releases core-matrix-update-20261106 (2026-11-06), logpower-ridge-update-20261106 (2026-11-06), foundation-checkpoints-update-20261106 (2026-11-06), eegpt-core-update-20261013 (2026-10-13), foundation-models-update-20261004 (2026-10-04), research-preview-20260920 (2026-09-20). Cite the upstream dataset as well: its credit is on this page.

Every release is archived on Zenodo: doi:10.5281/zenodo.23123296. BibTeX for the site and its releases → · CITATION.cff ↗