Core-matrix protocol · Transfer to a new person

Arithmetic & rest on EEGMAT

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

The protocol at a glance

Task
Serial subtraction versus resting EEG
Evaluation
Transfer to a new person
Dataset
EEGMAT
People
36
Data
2,160 epochs · 5 participant-disjoint folds
Input
19 channels · 2-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
56.8% (54.6%–58.9%)
LaBraM
64.6% (60.5%–68.6%)
CBraMod
62.3% (58.1%–66.5%)
EEGNet
67.6% (62.8%–72.5%)
TS+LR
65.5% (59.6%–71.8%)
MDM
71.5% (65.5%–77.8%)
ShallowFBCSPNet
67.9% (63.4%–72.4%)
Deep4Net
61.3% (57.5%–65.2%)
CSP+LDA
68.3% (62.0%–74.3%)
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 fit56.8%54.6%–58.9%0.558Mean across held-out participants19360.6 s
LaBraMFoundation modelFrozen encoder + ridge head64.6%60.5%–68.6%0.620Mean across held-out participants19362.1 s
CBraModFoundation modelFrozen encoder + ridge head62.3%58.1%–66.5%0.602Mean across held-out participants19366.1 s
EEGNetCompact modelScratch · 20 epochs67.6%62.8%–72.5%Single seed — the highest of the seeds run (see Stability)0.637Mean across held-out participants1936117.7 s
TS+LRClassical method · OAS · 4–40 Hz · primary row · Local Ubuntu / CPUSupervised fit65.5%59.6%–71.8%0.592Mean across held-out participants1936—
MDMClassical method · OAS · 4–40 Hz · secondary row · Local Ubuntu / CPUSupervised fit71.5%65.5%–77.8%0.661Mean across held-out participants1936—
ShallowFBCSPNetCompact model · matrix recipe · AdamW · Apple M5 / MPSScratch · 20 epochs67.9%63.4%–72.4%0.644Mean across held-out participants1936—
Deep4NetCompact model · matrix recipe · AdamW · Apple M5 / MPSScratch · 20 epochs61.3%57.5%–65.2%0.551Mean across held-out participants1936—
CSP+LDAClassical method · broadband CSP · shrinkage LDA · Apple M5 / CPUSupervised fit68.3%62.0%–74.3%0.649Mean across held-out participants1936—

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).

A frozen NeuroLM pilot with its own preprocessing is on the pilot studies page →

Frozen-encoder probes with their own preprocessing are on the pilot studies page: NeuroRVQ on three tasks → · EEG-Captioner and AXON on ten-second windows →

Pick-two protocol comparison

Compare any two methods on Arithmetic & rest

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%TS+LR65.5% [59.6–71.8]Spectral ridge56.8% [54.6–58.9]
Published paired contrastTS+LR − Spectral ridge: +8.8 pp[+1.6, +16.1 pp]

98.75% interval, Bonferroni-adjusted for four contrasts · n = 36 participants · Paired interval excludes zero

Default view (Arithmetic & rest): TS+LR balanced accuracy 65.5% (95% interval 59.6–71.8%); Spectral ridge 56.8% (95% interval 54.6–58.9%); chance 50.0%.

Published paired contrasts table (15 pairs)
Pair (left minus right)Mean differenceGoverning intervalParticipantsAdjustment & notes
TS+LR − Spectral ridge+8.8 pp[+1.6, +16.1 pp]3698.75% interval, Bonferroni-adjusted for four contrasts
TS+LR − EEGNet−2.1 pp[−8.0, +3.8 pp]36nominal 95% intervalagainst the matrix's one-seed EEGNet cell; training variance not included
MDM − EEGNet+3.8 pp[−1.3, +9.3 pp]36nominal 95% intervalagainst the matrix's one-seed EEGNet cell; training variance not included
MDM − Spectral ridge+14.7 pp[+8.9, +20.5 pp]36nominal 95% interval
TS+LR − LaBraM+0.9 pp[−4.6, +6.6 pp]36nominal 95% interval
TS+LR − CBraMod+3.2 pp[−1.5, +8.0 pp]36nominal 95% interval
TS+LR − MDM−6.0 pp[−11.6, −0.5 pp]36nominal 95% interval
CSP+LDA − EEGNet+0.7 pp[−4.3, +5.8 pp]3695% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
Deep4Net − EEGNet−6.3 pp[−10.4, −2.3 pp]3695% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
ShallowFBCSPNet − EEGNet+0.3 pp[−2.6, +3.3 pp]3695% interval, unadjustedagainst the matrix's one-seed EEGNet cell; training variance not included
ShallowFBCSPNet − Deep4Net+6.6 pp[+2.6, +10.6 pp]3695% interval
Log band-power ridge − Spectral ridge+8.2 pp[+4.0, +12.6 pp]3695% paired participant bootstrap interval, unadjusted
LaBraM at Scale Quark-L − LaBraM at Scale Quark-B−2.1 pp[−5.0, +0.8 pp]3695% paired participant bootstrap interval, unadjusted (21 ladder contrasts)
LaBraM at Scale Quark-H − LaBraM at Scale Quark-L+0.6 pp[−1.7, +2.8 pp]3695% paired participant bootstrap interval, unadjusted (21 ladder contrasts)
LaBraM at Scale Quark-H − LaBraM at Scale Quark-B−1.5 pp[−4.4, +1.3 pp]3695% paired participant bootstrap interval, unadjusted (21 ladder contrasts)

How these rows were run

Riemannian and xDAWN baselines

Recipe
TS+LR: OAS · 4–40 Hz · primary row; MDM: OAS · 4–40 Hz · secondary row
Hardware
TS+LR, MDM: 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.

The EEGMAT signals were ICA-cleaned and may be rank-deficient; OAS shrinkage then dominates the low-variance directions of the covariances that TS+LR and MDM read.

ShallowFBCSPNet, Deep4Net and CSP+LDA under the matrix recipe

Recipe
ShallowFBCSPNet, Deep4Net: matrix recipe · AdamW; CSP+LDA: broadband CSP · shrinkage LDA
Hardware
ShallowFBCSPNet, Deep4Net: Apple M5 / MPS; CSP+LDA: Apple M5 / 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)98.75% interval, Bonferroni-adjusted for four contrasts95% interval, unadjustedReading
TS+LR − Spectral ridgeTS+LR · OAS · 4–40 Hz+8.8 pp1.6 to 16.13.0 to 14.6higher

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+0.7 pp−4.3 to 5.8−7.1 to 8.7unresolved
Deep4Net − EEGNet−6.3 pp−10.4 to −2.3−12.8 to 0.1lower
ShallowFBCSPNet − EEGNet+0.3 pp−2.6 to 3.3−4.0 to 5.0unresolved

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.

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 fit56.76%54.58%–58.94%0.5581936
Log band-power ridge (5 bands, visible channels)Classical methodSupervised fit64.95%60.19%–69.72%0.6101936

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: +8.19 pp (+3.98 to +12.59): 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 14 frozen-encoder rows this comparison was specified against, chosen on these same test folds: −4.07 pp (−8.70 to +0.74): not separated. Choosing it on the test folds leans the comparison against this row.

Notes on this comparator

The band-power rows in the pilot studies panels on this cohort come from their own input chains and are read on their own.

Recording order on this protocol: the limits below ↓

On these windows and folds this is the same comparator as the third block of the panel on the components of Neural State Prediction on the pilot studies page →

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
56.8% (54.6%–58.9%)
LaBraM
64.6% (60.5%–68.6%)
CBraMod
62.3% (58.1%–66.5%)
EEGNet
67.6% (62.8%–72.5%)
REVE Base
68.8% (64.1%–73.2%)
REVE Large
69.0% (63.3%–74.7%)
LUNA Base
60.7% (56.9%–64.4%)
LUNA Large
57.4% (54.3%–60.6%)
BrainOmni Base
61.6% (58.4%–64.9%)
CodeBrain
63.0% (59.5%–66.5%)
EEGMamba
60.3% (57.0%–63.4%)
ST-EEGFormer Base
64.5% (59.0%–70.0%)
ST-EEGFormer Large
63.4% (57.8%–69.0%)
eeg-fm-masking MAE (r=9 cm, L=2)
63.6% (59.1%–68.1%)
ERP-FM Base
61.1% (57.3%–65.1%)
SingLEM
51.5% (49.2%–53.8%)
ZUNA 1.1
53.3% (49.5%–57.1%)
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 head68.8%64.1%–73.2%0.6611936not in the authors' published pretraining list (checked 2026-10-04)source ↗
REVE Large2Foundation modelFrozen encoder + ridge head69.0%63.3%–74.7%0.6611936not in the authors' published pretraining list (checked 2026-10-04)source ↗
LUNA Base3Foundation modelFrozen encoder + ridge head60.7%56.9%–64.4%0.5911936not in the authors' published pretraining list (checked 2026-10-04)source ↗
LUNA Large4Foundation modelFrozen encoder + ridge head57.4%54.3%–60.6%0.5611936not in the authors' published pretraining list (checked 2026-10-04)source ↗
BrainOmni Base5Foundation modelFrozen encoder + ridge head61.6%58.4%–64.9%0.6031936not in the authors' published pretraining list (checked 2026-10-04)source ↗
CodeBrain6Foundation modelFrozen encoder + ridge head63.0%59.5%–66.5%0.6091936not in the authors' published pretraining list (checked 2026-10-04)source ↗
EEGMamba7Foundation modelFrozen encoder + ridge head60.3%57.0%–63.4%0.5911936not in the authors' published pretraining list (checked 2026-10-04)source ↗
ST-EEGFormer Base8Foundation modelFrozen encoder + ridge head64.5%59.0%–70.0%0.6161936not in the authors' published pretraining list (checked 2026-10-04)source ↗
ST-EEGFormer Large9Foundation modelFrozen encoder + ridge head63.4%57.8%–69.0%0.5931936not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking MAE (r=9 cm, L=2)10Foundation modelFrozen encoder + ridge head63.6%59.1%–68.1%0.6101936not in the authors' published pretraining list (checked 2026-10-04)source ↗
ERP-FM Base11Foundation modelFrozen encoder + ridge head61.1%57.3%–65.1%0.5801936not in the authors' published pretraining list (checked 2026-10-04)source ↗
SingLEM12Foundation modelFrozen encoder + ridge head51.5%49.2%–53.8%Interval reaches chance level0.5141936not 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 head53.3%49.5%–57.1%Interval reaches chance level0.5101936Exposure 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 head63.6%59.1%–68.1%0.6101936not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking JEPA (r=9 cm, L=2)14Foundation modelFrozen encoder + ridge head61.6%56.4%–66.5%0.5881936not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking MAE (r=one channel, L=1)14Foundation modelFrozen encoder + ridge head64.3%60.2%–68.2%0.6231936not in the authors' published pretraining list (checked 2026-10-04)source ↗
eeg-fm-masking JEPA (r=one channel, L=1)14Foundation modelFrozen encoder + ridge head62.5%58.5%–66.2%0.6011936not in the authors' published pretraining list (checked 2026-10-04)source ↗

Adapted on this protocol: nine encoders, one fixed recipe

The published LaBraM adaptation recipe, unchanged: a head trained on the frozen encoder against rank-4 LoRA with the same head, trained for five epochs on the same participant-disjoint folds, over three seeds. The frozen ridge readout in the table above, the trained head and LoRA order the models differently here. 36 people, 5 folds. LoRA adds 32,768 to 428,032 trainable parameters, depending on the model.

EncoderFrozen encoder + trained headLoRA rank 4 + headLoRA minus frozen + headTrainable parameters (head · LoRA arm)
LaBraMLaBraM, already published56.6%54.2%–59.0%64.1%60.2%–67.8%+7.5 pp+4.8 to +10.1 pp30 improved · 6 worse · 0 unchangedThe interval excludes zero.402 · 38,802
REVE Base64.0%60.4%–67.7%66.3%61.9%–70.7%+2.3 pp+0.1 to +4.6 pp21 improved · 13 worse · 2 unchangedThe interval excludes zero.1,026 · 181,250
REVE Large65.6%61.1%–70.1%69.2%63.4%–74.9%+3.6 pp+0.5 to +6.8 pp21 improved · 13 worse · 2 unchangedThe interval excludes zero.2,434 · 430,466
LUNA Base58.2%54.5%–61.7%60.0%55.9%–63.9%+1.8 pp+0.1 to +3.3 pp23 improved · 11 worse · 2 unchangedThe interval excludes zero.514 · 37,378
BrainOmni Base64.5%59.6%–69.2%64.8%59.6%–69.7%+0.3 pp−1.4 to +1.8 pp18 improved · 14 worse · 4 unchangedThe interval includes zero: no change is established.16,386 · 106,498
CodeBrain66.5%62.3%–70.6%67.1%62.6%–71.6%+0.6 pp−0.7 to +1.9 pp18 improved · 14 worse · 4 unchangedThe interval includes zero: no change is established.7,602 · 58,802
EEGMamba60.4%55.7%–65.2%60.8%56.3%–65.3%+0.4 pp−0.7 to +1.5 pp18 improved · 14 worse · 4 unchangedThe interval includes zero: no change is established.7,602 · 90,930
ST-EEGFormer Base63.2%59.7%–66.7%65.1%61.0%–69.4%+1.9 pp−0.2 to +4.1 pp20 improved · 16 worse · 0 unchangedThe interval includes zero: no change is established.1,538 · 148,994
SingLEM50.9%49.5%–52.3%60.6%56.7%–64.4%+9.7 pp+6.1 to +13.3 pp30 improved · 6 worse · 0 unchangedThe interval excludes zero.1,218 · 33,986
ZUNA 1.1Exposure unknownModel card: research use only, not for diagnosis or clinical use.51.2%49.8%–52.7%56.9%53.7%–60.1%+5.7 pp+2.5 to +8.7 pp28 improved · 7 worse · 1 unchangedThe interval excludes zero.1,218 · 296,130

Not adapted: LUNA Large: not run by design: LUNA Large is frozen probes only in this evaluation’s specification; ST-EEGFormer Large: not run by design: ST-EEGFormer Large is frozen probes only; eeg-fm-masking MAE (r=9 cm, L=2): not run by design: eeg-fm-masking is a frozen-probe masking ablation; ERP-FM Base: not run by design: ERP-FM is frozen probes only and EEGMAT arithmetic is outside the ERP design; eeg-fm-masking JEPA (r=9 cm, L=2): not run by design: eeg-fm-masking is a frozen-probe masking ablation; eeg-fm-masking MAE (r=one channel, L=1): not run by design: eeg-fm-masking is a frozen-probe masking ablation; eeg-fm-masking JEPA (r=one channel, L=1): not run by design: eeg-fm-masking is a frozen-probe masking ablation

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
56.8% (54.6%–58.9%)
LaBraM
64.6% (60.5%–68.6%)
CBraMod
62.3% (58.1%–66.5%)
EEGNet
67.6% (62.8%–72.5%)
EEGPT
65.7% (60.9%–70.3%)
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 head65.7%60.9%–70.3%0.6371936not in the authors' published pretraining list (checked 2026-10-12)source ↗

Footnotes to this row

  1. 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.
  2. 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 / CPU66.0%62.0%–70.0%0.635—
Quark-LLocal Ubuntu / CPU63.8%59.8%–67.8%0.619—
Quark-HLocal Ubuntu / CPU64.5%60.6%–68.3%0.631—
Quark-G†Local Ubuntu / CUDA59.4%56.9%–61.7%0.586—
LaBraM at Scale · Atom-L and Molecule-L
Atom-LLocal Ubuntu / CPU67.5%63.7%–71.3%0.655its pretraining corpus includes this protocol's dataset
Molecule-LLocal Ubuntu / CPU63.9%59.9%–68.0%0.612its pretraining corpus includes this protocol's dataset
REVE Small · early checkpoint (epoch 15 of 200)
REVE SmallLocal Ubuntu / CPU65.2%60.5%–70.0%0.625—

† 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.

Two more models · this protocol’s people and folds

Signal-JEPA and EEGSimpleConv

Signal-JEPA as a frozen encoder with a ridge head fitted on the training people, and EEGSimpleConv trained from scratch for 20 epochs, both on this protocol’s windows, people and 5 folds. Each is compared with EEGNet on the same people.

ModelTraining modeInputBalanced accuracyMacro F1Difference from EEGNet
Signal-JEPAFoundation model · 3,456,852 parametersFrozen encoder + ridge head0.5–40 Hz band-pass, 128 Hz51.11%48.84%–53.47%Interval reaches chance level0.508−16.53 pp(−21.53 to −11.25)vs this page’s EEGNet cell: 67.6%against the matrix's one-seed EEGNet cell; training variance not included
EEGSimpleConvCompact model · 187,070 parametersTrained from scratch, 20 epochsthe matrix’s 250 Hz windows61.90%mean of three seeds0.577−5.62 ppvs EEGNet’s three-seed mean · no interval

Each difference is a paired comparison with EEGNet on the same people.

Notes on Signal-JEPA and EEGSimpleConv

Signal-JEPA’s input: each window band-passed 0.5–40 Hz and resampled to 128 Hz, the format it was pretrained on, in microvolts, recorded channels only. The further-encoder panel follows its own input contract, so this row sits apart from it.

Signal-JEPA’s difference is paired, on the same people, against this page’s EEGNet cell: 10,000 participant bootstrap draws, no multiplicity correction. EEGNet’s mean over its three seeds here is 67.52%.

EEGSimpleConv: the mean of three seeds, at width 64 and the matrix’s 20-epoch budget, fixed before scoring. Its difference is from EEGNet’s mean over the same three seeds and is printed without an interval.

Signal-JEPA’s model card names Lee2019 (OpenBMI) as its pretraining data.

Terms of Signal-JEPA and EEGSimpleConv

Limits

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.

Recording order: the PhysioNet record describes each person’s file ending _1 as rest recorded before the task and the file ending _2 as recorded during mental arithmetic, so recording order and elapsed time move with the label in every result on this protocol. PhysioNet record ↗

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. 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.
  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: pre-task-rest, mental-arithmetic.
  5. Uniform-guessing reference: 50.00%. Scores weight participants equally. Intervals describe participant variation; cross-validation training sets overlap.
  6. The benchmark detects condition (rest versus serial subtraction), not the good/bad count-quality participant grouping.
  7. Only the first documented 60 seconds is retained even though EDF containers are longer.
  8. The source README reports prior ICA artifact removal, so these are not untouched acquisition signals.
  9. Open Data Commons Attribution License v1.0 applies; retain PhysioNet attribution.
  10. 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.

Stability

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.

Seeds run for EEGNet: 67.64% · 67.55% · 67.36%; mean 67.52%

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.

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

EEGMAT

Credit 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.

Licence Open Data Commons Attribution License 1.0 ↗

Dataset record ↗ · Every result on this dataset →

The figures here are aggregate measurements computed by BCI Report under the licence above; the data remain their creators’.

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 Attribution license and study-specific approval/consent are documented; eligibility assumes aggregate-only output and exclusion of subject-info fields.

Rights reviewed · against physionet.org ↗ · www.mdpi.com ↗ · opendatacommons.org ↗

Correction on record · : The arithmetic-rest protocol’s limitations do not say that EEGMAT records rest before the task for every participant, as PhysioNet describes it. The protocol page states it beside the other limitations. Corrections register →

Reproducibility record

Protocol id
parallel-fixed-subject-folds-v1/physionet-eegmat-1.0.0
Dataset release
PhysioNet EEG During Mental Arithmetic Tasks 1.0.0 · 1.0.0
Hardware
Apple M5 / MPS
Scoring stage sum
126.5 s · may include accelerator waiting
Audit record SHA-256
2d0e4cec78c8d6e5bfc9f5390ba161f4308759e02feffbb8645d6983d9093292
Summary SHA-256
ab607e749da1eedcf88441c5b6fd0db55e2ebae0bbd91d6d18e508605eef625b
Protocol SHA-256
5d53ccbd5a00763908162f0b700c0a3236e0d4f9d43b08e42ee31a2ef01aefbf

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
Apple M5 / MPS · Apple M5 / 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). Arithmetic & rest on EEGMAT. https://bci.report/protocols/arithmetic-rest/

Figures from releases core-matrix-update-20261106 (2026-11-06), logpower-ridge-update-20261106 (2026-11-06), signal-jepa-eegsimpleconv-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 ↗