# Does pretraining help EEG foundation models like LaBraM and CBraMod?

Representation controls · matched frozen encoders

Sometimes. CBraMod and EEGPT gain under both readout settings on both tasks; LaBraM gains on mental workload and does not on MI/rest. Eleven further foundation models are compared in the same frozen recipe.

- [Reviewed aggregate JSON ↓](https://bci.report/data/deployment-topics.json)
- 

Measured on: [ds003810](https://bci.report/datasets/ds003810/); [EEGMAT](https://bci.report/datasets/eegmat/); [BETA](https://bci.report/datasets/beta/); [ds006593](https://bci.report/datasets/ds006593/); [TMNRED / ds005383](https://bci.report/datasets/tmnred/); [EESM19 scalp subset](https://bci.report/datasets/eesm19/) · Methods: [EEGNet](https://bci.report/methods/eegnet/); [LaBraM](https://bci.report/methods/labram/); [CBraMod](https://bci.report/methods/cbramod/); [CCA](https://bci.report/methods/cca/); [REVE](https://bci.report/methods/reve/); [LUNA](https://bci.report/methods/luna/); [BrainOmni](https://bci.report/methods/brainomni/); [CodeBrain](https://bci.report/methods/codebrain/); [EEGMamba](https://bci.report/methods/eegmamba/); [ST-EEGFormer](https://bci.report/methods/st-eegformer/); [eeg-fm-masking](https://bci.report/methods/eeg-fm-masking/); [ERP-FM](https://bci.report/methods/erp-fm/); [SingLEM](https://bci.report/methods/singlem/); [ZUNA](https://bci.report/methods/zuna/); [EEGPT](https://bci.report/methods/eegpt/)

## Short answer

Sometimes, depending on the encoder and the task: with frozen encoders and the fixed readout, pretrained CBraMod beat the mean of its three random initializations by **+6.39 pp** of balanced accuracy on MI/rest (**10** people) and by **+5.90 pp** on mental workload (**36** people), pretrained EEGPT beat its own by **+4.25 pp** and **+6.45 pp** on the same two tasks, and pretrained LaBraM beat its own by **+8.12 pp** on mental workload, each with a paired 95% interval above zero. On MI/rest, LaBraM's difference was **−2.42 pp** (interval crosses zero), and **−3.56 pp** under the train-selected readout (interval wholly below zero). Neither benchmark is in LaBraM's, CBraMod's or EEGPT's published pretraining list (checked 2026-10-04, EEGPT's 2026-10-12). The evidence is for these two tasks and frozen encoders. Eleven further foundation models were run in the same frozen recipe without a random-initialization control. On sleep staging (**20** people) REVE Large, ST-EEGFormer Base and ST-EEGFormer Large (**76.6%**, **77.3%**, **79.6%**) lie above every published row, spectral ridge at **72.2%** included, as does EEGPT, whose interval clears spectral ridge's by only **0.10 pp**; on BETA none lies above training-free CCA (**63.1%**); on EEGMAT the frozen readouts and LoRA order the encoders differently.

**Pretrained minus random: MI/rest, 10 people; mental workload, 36 people.** Difference in participant-mean balanced accuracy, in percentage points (pp): pretrained encoder minus the mean of three random initializations, same frozen architecture and splits. Dot: the difference; line: paired participant bootstrap 95% interval, printed under each label; dashed line: zero, no difference. Dark rows use the fixed readout, rust rows the train-selected readout.

- LaBraM · MI / rest — Fixed readout · −5.94 to +1.03 pp: −2.42 pp
- LaBraM · MI / rest — Train-selected readout · −7.00 to −0.19 pp: −3.56 pp
- CBraMod · MI / rest — Fixed readout · +3.06 to +9.50 pp: +6.39 pp
- CBraMod · MI / rest — Train-selected readout · +4.53 to +11.11 pp: +7.89 pp
- EEGPT · MI / rest — Fixed readout · +0.75 to +7.44 pp: +4.25 pp
- EEGPT · MI / rest — Train-selected readout · +0.72 to +7.08 pp: +3.92 pp
- LaBraM · Mental workload — Fixed readout · +4.97 to +11.40 pp: +8.12 pp
- LaBraM · Mental workload — Train-selected readout · +4.74 to +11.42 pp: +7.99 pp
- CBraMod · Mental workload — Fixed readout · +2.31 to +9.57 pp: +5.90 pp
- CBraMod · Mental workload — Train-selected readout · +3.12 to +10.43 pp: +6.73 pp
- EEGPT · Mental workload — Fixed readout · +2.61 to +10.59 pp: +6.45 pp
- EEGPT · Mental workload — Train-selected readout · +4.40 to +10.40 pp: +7.38 pp

Paired attribution contrasts

## Pretrained minus random, same readout

Each constructor-random architecture is averaged over three initialization seeds. Pretrained and random encoders share the same splits, pooling and ridge readout protocol.

Difference in participant-mean balanced accuracy. Intervals are paired participant bootstraps, conditional on three random initializations.

| Task / model | Fixed alpha 100 | Train-selected alpha |
| --- | --- | --- |
| LaBraM — MI / rest · n=10 | −2.42 pp (−5.94 to +1.03 pp) | −3.56 pp (−7.00 to −0.19 pp) |
| CBraMod — MI / rest · n=10 | +6.39 pp (+3.06 to +9.50 pp) | +7.89 pp (+4.53 to +11.11 pp) |
| EEGPT — MI / rest · n=10 | +4.25 pp (+0.75 to +7.44 pp) | +3.92 pp (+0.72 to +7.08 pp) |
| LaBraM — Mental workload · n=36 | +8.12 pp (+4.97 to +11.40 pp) | +7.99 pp (+4.74 to +11.42 pp) |
| CBraMod — Mental workload · n=36 | +5.90 pp (+2.31 to +9.57 pp) | +6.73 pp (+3.12 to +10.43 pp) |
| EEGPT — Mental workload · n=36 | +6.45 pp (+2.61 to +10.59 pp) | +7.38 pp (+4.40 to +10.40 pp) |

Measured scores

## Scores under both readout settings

Head selection uses only inner training participants.

Fixed head · alpha 100 · participant-mean balanced accuracy.

| Task / model | Pretrained encoder | Random encoder mean | Paired difference |
| --- | --- | --- | --- |
| LaBraM — MI / rest | 53.5% | 55.9% — 3 initializations | −2.42 pp (−5.94 to +1.03 pp) |
| CBraMod — MI / rest | 62.9% | 56.5% — 3 initializations | +6.39 pp (+3.06 to +9.50 pp) |
| EEGPT — MI / rest | 58.3% | 54.1% — 3 initializations | +4.25 pp (+0.75 to +7.44 pp) |
| LaBraM — Mental workload | 64.6% | 56.5% — 3 initializations | +8.12 pp (+4.97 to +11.40 pp) |
| CBraMod — Mental workload | 62.3% | 56.4% — 3 initializations | +5.90 pp (+2.31 to +9.57 pp) |
| EEGPT — Mental workload | 65.7% | 59.2% — 3 initializations | +6.45 pp (+2.61 to +10.59 pp) |

Train-selected head · alpha 100 / 10,000 / 1 · participant-mean balanced accuracy.

| Task / model | Pretrained encoder | Random encoder mean | Paired difference |
| --- | --- | --- | --- |
| LaBraM — MI / rest | 52.9% | 56.5% — 3 initializations | −3.56 pp (−7.00 to −0.19 pp) |
| CBraMod — MI / rest | 62.7% | 54.9% — 3 initializations | +7.89 pp (+4.53 to +11.11 pp) |
| EEGPT — MI / rest | 57.0% | 53.1% — 3 initializations | +3.92 pp (+0.72 to +7.08 pp) |
| LaBraM — Mental workload | 64.4% | 56.5% — 3 initializations | +7.99 pp (+4.74 to +11.42 pp) |
| CBraMod — Mental workload | 68.4% | 61.7% — 3 initializations | +6.73 pp (+3.12 to +10.43 pp) |
| EEGPT — Mental workload | 66.0% | 58.6% — 3 initializations | +7.38 pp (+4.40 to +10.40 pp) |

EEGPT’s rows: its pretrained rows under the fixed head are its core-matrix cells on these two protocols, the same predictions, so the notes on those cells hold here. 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. ds003810 is recorded at 125 Hz; its windows reach the encoder upsampled to 256 Hz, so they carry nothing above 62.5 Hz.

Fixed classical control

52.3%

Log-covariance ridge · MI / rest

— 95% interval 48.5%–56.4%

Fixed classical control

61.9%

Log-covariance ridge · Mental workload

— 95% interval 56.2%–67.6%

Training-seed sensitivity

## EEGNet across three training seeds

Five existing EEGNet protocols were rerun with three fixed MPS seeds. The range is a small-sample check of training sensitivity, not a population interval; the published point estimates stand.

Balanced accuracy across three training seeds. Sample SD and full range are percentage points.

| Protocol | Three runs | Mean | Sample SD | Full range |
| --- | --- | --- | --- | --- |
| Semantic ERP — n=30 · 3,600 trials | 60.50% / 60.42% / 60.33% | 60.42% | 0.08 pp | 0.17 pp — 60.33%–60.50% |
| P300 — n=21 · 2,520 trials | 55.79% / 50.36% / 52.06% | 52.74% | 2.78 pp | 5.44 pp — 50.36%–55.79% |
| Scalp sleep staging — n=20 · 3,000 trials | 57.33% / 54.43% / 51.60% | 54.46% | 2.87 pp | 5.73 pp — 51.60%–57.33% |
| BETA · 4 selected channels — n=70 · 11,200 trials | 44.77% / 44.42% / 45.03% | 44.74% | 0.30 pp | 0.61 pp — 44.42%–45.03% |
| BETA · 8 selected channels — n=70 · 11,200 trials | 56.54% / 56.73% / 56.99% | 56.75% | 0.23 pp | 0.46 pp — 56.54%–56.99% |

Dataset authors, versions, licenses and source records for all four seed-sensitivity datasets are listed under [Data use & privacy → Core and deployment sources](https://bci.report/data-use/#sources).

Frozen-probe comparison

## Further foundation encoders, frozen

Eleven further EEG foundation models — sixteen encoder checkpoints — were swapped into the published LaBraM and CBraMod frozen recipe, with the same windows, people, folds, ridge head and scoring, and no random-initialization control.

EEGPT sits in the sleep and BETA tables too, from its own release: it ran in the same runs as the published LaBraM and CBraMod rows with only the encoder swapped, and its random-initialization control is in the sections above.

A row is above or below another only where their 95% intervals do not overlap; with 126 frozen cells from the further encoders and EEGPT’s eight more, an occasional non-overlap is expected by chance.

### Sleep staging: four new encoders above every published row

On the core matrix’s sleep protocol (EESM19 scalp subset, 20 people, six channels, chance 20.0%), REVE Large, ST-EEGFormer Base and ST-EEGFormer Large lie entirely above every published row — spectral ridge, the published non-foundation row with the highest point estimate (72.2%, 70.7%–73.8%), the frozen LaBraM and CBraMod rows, and EEGNet. So does EEGPT, by a narrow margin: the lower end of its interval, 73.9%, clears the upper end of spectral ridge’s by 0.10 pp. Sleep is the only one of the seven scored protocols where a new row clears the published non-foundation row with the highest point estimate. The other new rows overlap the published ones or lie below them.

**Published rows on sleep staging (core matrix):** 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%)

Sleep staging · frozen encoder and ridge head · balanced accuracy with its descriptive 95% interval.

| Encoder | Balanced accuracy | Against LaBraM, CBraMod and spectral ridge |
| --- | --- | --- |
| [REVE Base](https://bci.report/methods/reve/) | 74.6% (72.9%–76.4%) | Overlaps all three |
| [REVE Large](https://bci.report/methods/reve/) | 76.6% (75.2%–78.0%) | Above all three |
| [LUNA Base](https://bci.report/methods/luna/) | 67.5% (65.3%–69.5%) | Overlaps LaBraM and CBraMod; below spectral ridge |
| [LUNA Large](https://bci.report/methods/luna/) | 72.4% (70.4%–74.0%) | Overlaps all three |
| [BrainOmni Base](https://bci.report/methods/brainomni/) | 67.7% (65.8%–69.5%) | Overlaps LaBraM and CBraMod; below spectral ridge |
| [CodeBrain](https://bci.report/methods/codebrain/) | 70.7% (68.8%–72.7%) | Overlaps all three |
| [EEGMamba](https://bci.report/methods/eegmamba/) | 64.1% (61.6%–66.4%) | Below all three |
| [ST-EEGFormer Base](https://bci.report/methods/st-eegformer/) | 77.3% (75.6%–78.8%) | Above all three |
| [ST-EEGFormer Large](https://bci.report/methods/st-eegformer/) | 79.6% (78.2%–81.0%) | Above all three |
| [eeg-fm-masking MAE (r=9 cm, L=2)](https://bci.report/methods/eeg-fm-masking/) | 67.9% (65.4%–70.4%) | Overlaps LaBraM and CBraMod; below spectral ridge |
| [ERP-FM Base](https://bci.report/methods/erp-fm/) | 71.0% (69.5%–72.6%) | Overlaps all three |
| [SingLEM](https://bci.report/methods/singlem/) | 35.3% (33.7%–37.0%) | Below all three |
| [ZUNA 1.1](https://bci.report/methods/zuna/) — Model card: research use only, not for diagnosis or clinical use. | 60.1% (57.8%–62.3%) Exposure unknown | Below all three |
| [EEGPT](https://bci.report/methods/eegpt/) — Same runs as the published rows | 75.5% (73.9%–77.1%) | Above LaBraM (intervals 0.67 pp apart), CBraMod (intervals 0.17 pp apart) and spectral ridge (intervals 0.10 pp apart) |

Against the published rows: LaBraM, CBraMod and spectral ridge, the published non-foundation row with the highest point estimate, by whether the 95% intervals overlap. Unless a row is marked, its dataset is not in the model authors’ published pretraining list (checked 2026-10-04); each model’s sources are on its method page. EEGPT’s row gives the gap between the two intervals beside each relation; its authors’ list was checked on 2026-10-12.

### BETA: above CBraMod, overlapping or below training-free CCA

On BETA’s 40-target SSVEP with eight electrodes (70 people, chance 2.5%), REVE Large and eeg-fm-masking MAE (r=9 cm, L=2) lie above the published frozen CBraMod row (33.7%), as do the three masking-ablation siblings shown below. No new row lies above standard CCA (63.1%), which needs no training at all. ST-EEGFormer’s BETA cells are in its authors’ pretraining list, and they still lie below CBraMod.

**Published rows on BETA at eight electrodes (core matrix):** Standard CCA 63.1% (57.2%–69.0%) · Spectral ridge 50.7% (45.6%–55.8%) · LaBraM 10.8% (9.5%–12.2%) · CBraMod 33.7% (29.8%–37.6%) · EEGNet 55.8% (50.1%–61.3%)

SSVEP, BETA, eight electrodes · frozen encoder and ridge head · balanced accuracy with its descriptive 95% interval.

| Encoder | Balanced accuracy | Against LaBraM, CBraMod and standard CCA |
| --- | --- | --- |
| [REVE Base](https://bci.report/methods/reve/) | 37.5% (33.3%–41.9%) | Above LaBraM; overlaps CBraMod; below standard CCA |
| [REVE Large](https://bci.report/methods/reve/) | 52.7% (47.7%–57.8%) | Above LaBraM and CBraMod; overlaps standard CCA |
| [LUNA Base](https://bci.report/methods/luna/) | 8.6% (7.7%–9.5%) | Below all three |
| [LUNA Large](https://bci.report/methods/luna/) | 10.0% (8.9%–11.1%) | Overlaps LaBraM; below CBraMod and standard CCA |
| [BrainOmni Base](https://bci.report/methods/brainomni/) | 21.7% (18.6%–25.0%) | Above LaBraM; below CBraMod and standard CCA |
| [CodeBrain](https://bci.report/methods/codebrain/) | 11.9% (10.5%–13.3%) | Overlaps LaBraM; below CBraMod and standard CCA |
| [EEGMamba](https://bci.report/methods/eegmamba/) | 40.2% (35.7%–44.7%) | Above LaBraM; overlaps CBraMod; below standard CCA |
| [ST-EEGFormer Base](https://bci.report/methods/st-eegformer/) | 17.1% (15.5%–18.8%) in the authors' published pretraining list | Above LaBraM; below CBraMod and standard CCA |
| [ST-EEGFormer Large](https://bci.report/methods/st-eegformer/) | 22.6% (20.1%–25.3%) in the authors' published pretraining list | Above LaBraM; below CBraMod and standard CCA |
| [eeg-fm-masking MAE (r=9 cm, L=2)](https://bci.report/methods/eeg-fm-masking/) | 54.8% (49.4%–60.0%) | Above LaBraM and CBraMod; overlaps standard CCA |
| [ERP-FM Base](https://bci.report/methods/erp-fm/) | 18.4% (16.3%–20.5%) | Above LaBraM; below CBraMod and standard CCA |
| [SingLEM](https://bci.report/methods/singlem/) | 26.6% (23.7%–29.5%) | Above LaBraM; below CBraMod and standard CCA |
| [ZUNA 1.1](https://bci.report/methods/zuna/) — Model card: research use only, not for diagnosis or clinical use. | 11.9% (10.8%–13.0%) Exposure unknown | Overlaps LaBraM; below CBraMod and standard CCA |
| [EEGPT](https://bci.report/methods/eegpt/) — Same runs as the published rows | 20.3% (18.2%–22.6%) | Above LaBraM (intervals 5.97 pp apart); below CBraMod (intervals 7.24 pp apart) and standard CCA (intervals 34.64 pp apart) |

Against the published rows: LaBraM, CBraMod and standard CCA, the published non-foundation row with the highest point estimate (its interval overlaps EEGNet’s), by whether the 95% intervals overlap. Unless a row is marked, its dataset is not in the model authors’ published pretraining list (checked 2026-10-04); each model’s sources are on its method page. EEGPT’s row gives the gap between the two intervals beside each relation; its authors’ list was checked on 2026-10-12.

**EEGPT on BETA:** PO5 and PO6 use channel embeddings that EEGPT’s pretraining never trained: they are two of the four names in its 62-name channel dictionary (AF7, AF8, PO5, PO6) that its 58-electrode pretraining montage leaves out. The other six channels here, and all four of the 4-channel subset, use trained embeddings. For context: BETA is not in EEGPT’s published pretraining list, but the Tsinghua 40-target SSVEP benchmark (Wang et al., 2017) is, as TSU. BETA’s authors describe BETA as an extension of that benchmark, from the same group with the same 40-target layout and its own recordings; whether anyone took part in both is not stated.

### EEGMAT: frozen readouts and LoRA disagree

Nine of the new encoders were also adapted on EEGMAT, mental arithmetic against rest (36 people, 5 participant-disjoint folds, three seeds, chance 50.0%), with the recipe of the published LaBraM adaptation: a head trained on the frozen encoder against rank-4 LoRA with the same head, five epochs, no tuning. LoRA’s budget differs by architecture, from 32,768 to 428,032 added trainable parameters.

The ridge readout of the frozen-probe table, the trained head and LoRA do not put the encoders in the same order. LoRA’s change excludes zero for REVE Base, REVE Large, LUNA Base, SingLEM and ZUNA 1.1; it includes zero for BrainOmni Base, CodeBrain, EEGMamba and ST-EEGFormer Base.

The clearest case is SingLEM: both frozen readouts reach chance (51.5% with the ridge head, 50.9% with the trained head), while LoRA reached 60.6%.

EEGMAT · balanced accuracy with its descriptive 95% interval · the paired change over the same people and folds, with how many people moved which way · trainable parameters of the head alone and of the LoRA arm.

| Encoder | Frozen, ridge head (frozen-probe table) | Frozen, trained head | LoRA rank 4 + head | LoRA minus trained head | Trainable parameters · head · LoRA arm |
| --- | --- | --- | --- | --- | --- |
| [LaBraM](https://bci.report/methods/labram/) — LaBraM, already published | 64.6% (60.5%–68.6%) | 56.6% (54.2%–59.0%) | 64.1% (60.2%–67.8%) | +7.5 pp (+4.8 to +10.1 pp) — 30 improved · 6 worse · 0 unchanged The interval excludes zero. | 402 · 38,802 |
| [REVE Base](https://bci.report/methods/reve/) | 68.8% (64.1%–73.2%) | 64.0% (60.4%–67.7%) | 66.3% (61.9%–70.7%) | +2.3 pp (+0.1 to +4.6 pp) — 21 improved · 13 worse · 2 unchanged The interval excludes zero. | 1,026 · 181,250 |
| [REVE Large](https://bci.report/methods/reve/) | 69.0% (63.3%–74.7%) | 65.6% (61.1%–70.1%) | 69.2% (63.4%–74.9%) | +3.6 pp (+0.5 to +6.8 pp) — 21 improved · 13 worse · 2 unchanged The interval excludes zero. | 2,434 · 430,466 |
| [LUNA Base](https://bci.report/methods/luna/) | 60.7% (56.9%–64.4%) | 58.2% (54.5%–61.7%) | 60.0% (55.9%–63.9%) | +1.8 pp (+0.1 to +3.3 pp) — 23 improved · 11 worse · 2 unchanged The interval excludes zero. | 514 · 37,378 |
| [BrainOmni Base](https://bci.report/methods/brainomni/) | 61.6% (58.4%–64.9%) | 64.5% (59.6%–69.2%) | 64.8% (59.6%–69.7%) | +0.3 pp (−1.4 to +1.8 pp) — 18 improved · 14 worse · 4 unchanged The interval includes zero: no change is established. | 16,386 · 106,498 |
| [CodeBrain](https://bci.report/methods/codebrain/) | 63.0% (59.5%–66.5%) | 66.5% (62.3%–70.6%) | 67.1% (62.6%–71.6%) | +0.6 pp (−0.7 to +1.9 pp) — 18 improved · 14 worse · 4 unchanged The interval includes zero: no change is established. | 7,602 · 58,802 |
| [EEGMamba](https://bci.report/methods/eegmamba/) | 60.3% (57.0%–63.4%) | 60.4% (55.7%–65.2%) | 60.8% (56.3%–65.3%) | +0.4 pp (−0.7 to +1.5 pp) — 18 improved · 14 worse · 4 unchanged The interval includes zero: no change is established. | 7,602 · 90,930 |
| [ST-EEGFormer Base](https://bci.report/methods/st-eegformer/) | 64.5% (59.0%–70.0%) | 63.2% (59.7%–66.7%) | 65.1% (61.0%–69.4%) | +1.9 pp (−0.2 to +4.1 pp) — 20 improved · 16 worse · 0 unchanged The interval includes zero: no change is established. | 1,538 · 148,994 |
| [SingLEM](https://bci.report/methods/singlem/) | 51.5% (49.2%–53.8%) Interval reaches chance level | 50.9% (49.5%–52.3%) Interval reaches chance level | 60.6% (56.7%–64.4%) | +9.7 pp (+6.1 to +13.3 pp) — 30 improved · 6 worse · 0 unchanged The interval excludes zero. | 1,218 · 33,986 |
| [ZUNA 1.1](https://bci.report/methods/zuna/) Exposure unknown — Model card: research use only, not for diagnosis or clinical use. | 53.3% (49.5%–57.1%) Interval reaches chance level | 51.2% (49.8%–52.7%) Interval reaches chance level | 56.9% (53.7%–60.1%) | +5.7 pp (+2.5 to +8.7 pp) — 28 improved · 7 worse · 1 unchanged The interval excludes zero. | 1,218 · 296,130 |

**The others were 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), eeg-fm-masking JEPA (r=9 cm, L=2), eeg-fm-masking MAE (r=one channel, L=1) and eeg-fm-masking JEPA (r=one channel, L=1) (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).

### REVE Base and Large: a size comparison

REVE Large’s point estimate is at or above REVE Base’s on all seven scored protocols; their intervals do not overlap on BETA at eight electrodes, BETA at four electrodes, the P300 target ERP and the semantic target ERP. Both checkpoints were scored on the same people, but each interval is marginal and no paired test was run, so a gap is a description, not an established effect of size.

REVE Base and Large on every scored protocol: the table

Frozen encoder and ridge head · balanced accuracy with its descriptive 95% interval · Large minus Base as the export states it, and whether the two intervals overlap.

| Protocol | REVE Base | REVE Large | Large minus Base | Intervals |
| --- | --- | --- | --- | --- |
| [Motor imagery & rest](https://bci.report/protocols/mi-rest/#foundation-v9) — 10 people | 66.2% (61.8%–70.6%) | 70.2% (64.0%–76.1%) | +4.0 pp | Overlap |
| [SSVEP · 8 channels](https://bci.report/protocols/beta-8ch/#foundation-v9) — 70 people | 37.5% (33.3%–41.9%) | 52.7% (47.7%–57.8%) | +15.2 pp | **Do not overlap** |
| [SSVEP · 4 channels](https://bci.report/protocols/beta-4ch/#foundation-v9) — 70 people | 25.5% (22.4%–28.8%) | 46.4% (41.7%–51.2%) | +20.8 pp | **Do not overlap** |
| [Arithmetic & rest](https://bci.report/protocols/arithmetic-rest/#foundation-v9) — 36 people | 68.8% (64.1%–73.2%) | 69.0% (63.3%–74.7%) | +0.2 pp | Overlap |
| [P300 target ERP](https://bci.report/protocols/p300-target/#foundation-v9) — 21 people | 50.5% (48.5%–52.3%) Interval reaches chance level | 54.6% (52.7%–56.5%) | +4.1 pp | **Do not overlap** |
| [Semantic target ERP](https://bci.report/protocols/semantic-target/#foundation-v9) — 30 people | 53.7% (51.8%–55.4%) | 59.7% (57.5%–61.8%) | +6.0 pp | **Do not overlap** |
| [Sleep staging](https://bci.report/protocols/sleep-scalp/#foundation-v9) — 20 people | 74.6% (72.9%–76.4%) | 76.6% (75.2%–78.0%) | +2.0 pp | Overlap |

The other two size pairs do not show this pattern: LUNA’s Base and Large intervals separate only on sleep staging, ST-EEGFormer’s only on BETA at eight electrodes and BETA at four electrodes, where its cells are in the authors’ pretraining list.

### Masking ablation: not separable on six of seven protocols

eeg-fm-masking released four checkpoints of one study: masking framework (MAE or JEPA) crossed with mask geometry (the paper-recommended r = 9 cm with L = 2, or a single channel with L = 1), with the same architecture, corpus and recipe. The paper-recommended MAE checkpoint is a matrix row in the tables above; the other three are its ablation. On six of the seven scored protocols every pair of the four overlaps. On sleep staging one pair separates — JEPA at r = 9 cm against MAE on one channel — and it changes the framework and the geometry at once, so it isolates neither.

The four eeg-fm-masking checkpoints on every scored protocol: the table

Frozen encoder and ridge head · balanced accuracy with its descriptive 95% interval.

| Protocol | MAE · r = 9 cm, L = 2 (matrix row) | JEPA · r = 9 cm, L = 2 | MAE · one channel, L = 1 | JEPA · one channel, L = 1 | Pairs whose intervals separate |
| --- | --- | --- | --- | --- | --- |
| [Motor imagery & rest](https://bci.report/protocols/mi-rest/#foundation-v9) | 68.1% (64.3%–72.3%) | 63.2% (58.2%–67.9%) | 64.9% (60.3%–69.4%) | 63.3% (59.0%–67.4%) | None: every pair overlaps |
| [SSVEP · 8 channels](https://bci.report/protocols/beta-8ch/#foundation-v9) | 54.8% (49.4%–60.0%) | 49.6% (44.5%–54.8%) | 53.4% (48.1%–58.6%) | 51.0% (45.8%–56.1%) | None: every pair overlaps |
| [SSVEP · 4 channels](https://bci.report/protocols/beta-4ch/#foundation-v9) | 47.1% (42.1%–52.1%) | 43.4% (38.7%–48.3%) | 46.7% (41.9%–51.5%) | 45.4% (40.6%–50.2%) | None: every pair overlaps |
| [Arithmetic & rest](https://bci.report/protocols/arithmetic-rest/#foundation-v9) | 63.6% (59.1%–68.1%) | 61.6% (56.4%–66.5%) | 64.3% (60.2%–68.2%) | 62.5% (58.5%–66.2%) | None: every pair overlaps |
| [P300 target ERP](https://bci.report/protocols/p300-target/#foundation-v9) | 49.5% (48.1%–51.0%) At or below chance level | 51.6% (50.4%–52.7%) | 51.1% (49.2%–53.0%) Interval reaches chance level | 50.0% (48.7%–51.5%) Interval reaches chance level | None: every pair overlaps |
| [Semantic target ERP](https://bci.report/protocols/semantic-target/#foundation-v9) | 54.8% (52.7%–56.9%) | 55.8% (53.9%–57.6%) | 54.9% (53.4%–56.4%) | 55.0% (53.4%–56.6%) | None: every pair overlaps |
| [Sleep staging](https://bci.report/protocols/sleep-scalp/#foundation-v9) | 67.9% (65.4%–70.4%) | 70.3% (68.0%–72.7%) | 65.2% (63.0%–67.3%) | 68.3% (65.7%–70.8%) | **JEPA at r = 9 cm vs. MAE on one channel** |

### Limits of the further encoders

### One fixed recipe, not a ceiling

The frozen probe is the published linear head with only the encoder swapped; the adaptation is one untuned five-epoch recipe on one task.

### Small cohorts, descriptive intervals

Motor imagery has 10 people, P300 21, sleep staging 20. The intervals are participant bootstraps that ignore cross-validation dependence and carry no multiplicity correction.

### Family order

Rows keep the export’s family order; the same people under two encoders are not tested as a pair.

### Method comparisons

Balanced or class-matched designs (P300, semantic target, motor imagery against rest, sampled sleep epochs) compare methods; they are not detection, false-alarm or latency estimates for real use.

### New people, same task and setup

The folds hold out people. Nothing here is cross-day, cross-device or cross-dataset evidence, and nothing is a clinical claim.

### Short windows

Every encoder sees the published 1–2 s windows (sleep as 15 two-second segments), usually shorter than its pretraining context; each model’s input handling is in its row footnote on the protocol pages.

### Pretraining exposure: the authors’ lists

BETA is in ST-EEGFormer’s published pretraining list and TMNRED in SingLEM’s; ZUNA 1.1 publishes none, so every ZUNA 1.1 cell is unknown; EEGMamba’s list is read from its official code, with medium confidence. Every other cell, LaBraM’s and CBraMod’s included, is not in the authors’ published pretraining list (checked 2026-10-04), and neither are EEGPT’s (checked 2026-10-12) — not proof that the recordings were never seen.

### No endorsement

No model’s authors endorse these results. No adapted weights or LoRA deltas were kept or shared, and each model’s weights licence travels with its rows. Timings were measured on a shared GPU and are not published.

### 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 ↗](https://arxiv.org/abs/2510.21585)
- **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 ↗](https://arxiv.org/abs/2510.22257)
- **BrainOmni Base**: MIT [paper ↗](https://arxiv.org/abs/2505.18185)
- **CodeBrain**: Apache-2.0 [paper ↗](https://arxiv.org/abs/2506.09110)
- **EEGMamba**: MIT [paper ↗](https://doi.org/10.1016/j.neunet.2025.107816)
- **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 ↗](https://openreview.net/forum?id=5Xwm8e6vbh)
- **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 ↗](https://arxiv.org/abs/2609.33487)
- **ERP-FM Base**: CC BY-NC-SA 4.0 (non-commercial; aggregate scores only) CC BY-NC-SA 4.0: non-commercial; aggregate scores only; attribution required; no weights redistributed. [paper ↗](https://arxiv.org/abs/2609.32796)
- **SingLEM**: MIT (checkpoint committed in the MIT repository) MIT. [paper ↗](https://arxiv.org/abs/2509.17920)
- **ZUNA 1.1**: Apache-2.0; model card: research use only, not for diagnosis or clinical use Apache-2.0 weights; model card states research use only, not for diagnosis or clinical use. [paper ↗](https://arxiv.org/abs/2607.27308)

EEGPT: the checkpoint’s licence is unresolved, and no original or converted weights are distributed; Apache-2.0 covers the repository’s code only.

Every row, with its footnote and exposure source, is on its protocol page ([Motor imagery & rest](https://bci.report/protocols/mi-rest/#foundation-v9), [SSVEP · 8 channels](https://bci.report/protocols/beta-8ch/#foundation-v9), [SSVEP · 4 channels](https://bci.report/protocols/beta-4ch/#foundation-v9), [Arithmetic & rest](https://bci.report/protocols/arithmetic-rest/#foundation-v9), [P300 target ERP](https://bci.report/protocols/p300-target/#foundation-v9), [Semantic target ERP](https://bci.report/protocols/semantic-target/#foundation-v9), [Sleep staging](https://bci.report/protocols/sleep-scalp/#foundation-v9)); the adaptation arms and the comparisons above are in [foundation-models-update.json](https://bci.report/data/foundation-models-update.json). Release [`foundation-models-update-20261004`](https://bci.report/releases/#foundation-models-update-20261004).

EEGPT’s rows are in their own section of the same protocol pages, from [eegpt-core-update.json](https://bci.report/data/eegpt-core-update.json), release [`eegpt-core-update-20261013`](https://bci.report/releases/#eegpt-core-update-20261013).

Methods & limits

## Limits of the pretraining contrasts

### Fixed versus selected heads

The primary head fixes ridge alpha at 100. The sensitivity setting selects among 100, 10,000 and 1 using inner training participants only.

### Three seeds, one cohort

The random baseline averages three encoder initializations. The participant bootstrap does not include unrestricted retraining uncertainty.

### Task-specific evidence

MI/rest contains 10 participants and 1,200 trials; mental workload contains 36 participants and 2,160 trials.

### Exposure of the two benchmarks

Neither ds003810 nor EEGMAT is in LaBraM’s or CBraMod’s published pretraining list, checked on 4 October 2026, or in EEGPT’s, checked on 12 October 2026 (sources below).

Pretraining lists, checked 2026-10-04: [LaBraM ↗](https://arxiv.org/abs/2405.18765) · [CBraMod ↗](https://arxiv.org/abs/2412.07236) · EEGPT’s, checked 2026-10-12: [EEGPT ↗](https://papers.neurips.cc/paper_files/paper/2024/file/4540d267eeec4e5dbd9dae9448f0b739-Paper-Conference.pdf) · [EEGPT repository ↗](https://github.com/BINE022/EEGPT/tree/ea7282f698c406cd423ccbedbe9960853761144c/datasets/pretrain)

Frozen CBraMod with a ridge readout is also read on a later session of the same person on WBCIC-SHU, without a random-initialization control: [Does a decoder trained on an earlier session still work later?](https://bci.report/topics/later-sessions/#wbcic-shu)

Frozen EEGPT with a supervised probe is read against matched classical features in one small pilot, on recorded orthosis device errors, without a random-initialization control: [What can a small pilot show, and what can it not?](https://bci.report/topics/pilot-studies/#orthosis)

### MI / rest · OpenNeuro ds003810 v2.0.2

Peterson et al. · CC0-1.0.

[Dataset record ↗](https://doi.org/10.18112/openneuro.ds003810.v2.0.2)

### Mental workload · EEGMAT 1.0.0

Zyma et al. · Open Data Commons Attribution License 1.0.

[PhysioNet record ↗](https://physionet.org/content/eegmat/1.0.0/)

Data source: [reviewed aggregate JSON](https://bci.report/data/deployment-topics.json) · schema bci-report-public-deployment-topics-v1.

EEGPT’s rows: [eegpt-core-update.json](https://bci.report/data/eegpt-core-update.json) · schema bci-report-eegpt-core-update-v1.

Keep exploring

Transfer

- [— Sensor transfer **Dry vs. wet electrodes**](https://bci.report/topics/dry-vs-wet/)
- [— Context transfer **Screen to VR**](https://bci.report/topics/screen-to-vr/)
- [— Montage **Fewer electrodes**](https://bci.report/topics/fewer-electrodes/)
- [— Motion robustness **On the move**](https://bci.report/topics/on-the-move/)
- [— Session transfer · same person **Later sessions**](https://bci.report/topics/later-sessions/)

Adapting models

- [— Calibration budget **How much calibration?**](https://bci.report/topics/calibration-budget/)
- [— Model adaptation **Which part to update?**](https://bci.report/topics/model-adaptation/)
- [— Representation controls **Does pretraining help?**](https://bci.report/topics/does-pretraining-help/)
- [— Shared encoder · route 2 **One model, several questions**](https://bci.report/topics/shared-encoder/)
- [— Questions in language · route 3 **Questions in language**](https://bci.report/topics/questions-in-language/)
- [— Route 3 follow-ups **New SSVEP frequencies**](https://bci.report/topics/new-ssvep-frequencies/)

Reliability & clinical

- [— Abstention · decision research **When not to act**](https://bci.report/topics/when-not-to-act/)
- [— Sleep staging · simple baselines **Sleep-stage balance**](https://bci.report/topics/sleep-staging/)
- [— Clinical research **Clinical groups**](https://bci.report/topics/clinical-groups/)
- [— Data quality and reproducibility **Checked end to end**](https://bci.report/topics/data-quality/)

Other signals & pilots

- [— Other signals · human and macaque **Beyond scalp EEG**](https://bci.report/topics/beyond-scalp-eeg/)
- [— Pilot studies · small cohorts **Pilot studies**](https://bci.report/topics/pilot-studies/)

[All questions and the transfer-coverage map →](https://bci.report/topics/)

## Cite this page

BCI Report (2026). *Does pretraining help EEG foundation models like LaBraM and CBraMod?* https://bci.report/topics/does-pretraining-help/

Figures from releases [`eegpt-core-update-20261013`](https://bci.report/releases/#eegpt-core-update-20261013) (2026-10-13), [`foundation-models-update-20261004`](https://bci.report/releases/#foundation-models-update-20261004) (2026-10-04), [`deployment-topics-20260920-v1`](https://bci.report/releases/#deployment-topics-20260920-v1) (2026-09-20), [`research-preview-20260920`](https://bci.report/releases/#research-preview-20260920) (2026-09-20). Cite the upstream datasets as well: their credits are on this page.

Every release is archived on Zenodo: [doi:10.5281/zenodo.23123296](https://doi.org/10.5281/zenodo.23123296). [BibTeX for the site and its releases →](https://bci.report/api/#cite-heading) · [CITATION.cff ↗](https://raw.githubusercontent.com/twu3202/bci-report/main/CITATION.cff)

---
Markdown copy of https://bci.report/topics/does-pretraining-help/, generated from the published page. Figures are aggregate results; terms of use: https://bci.report/data-use/