Fixed classical control
52.3%
52.3% balanced accuracyLog-covariance ridge · MI / rest
95% interval 48.5%–56.4%Representation controls · matched frozen encoders
Does pretraining help?
Sometimes. CBraMod gains under both readout settings on both tasks; LaBraM gains on mental workload and does not on MI/rest. These are matched frozen-encoder results, conditional on the tasks, splits and three constructor-random initializations—not a universal model ranking.
Paired attribution contrasts
Pretrained minus random, with the head held honest
Each constructor-random architecture is averaged over three initialization seeds. Pretrained and random encoders share the same splits, pooling and ridge readout protocol.
| Task / model | Fixed alpha 100 | Train-selected alpha |
|---|---|---|
| LaBraMMI / rest · n=10 | −2.42 pp−5.94 to +1.03 pp−2.42 pp pretraining contrast | −3.56 pp−7.00 to −0.19 pp−3.56 pp pretraining contrast |
| CBraModMI / rest · n=10 | +6.39 pp+3.06 to +9.50 pp+6.39 pp pretraining contrast | +7.89 pp+4.53 to +11.11 pp+7.89 pp pretraining contrast |
| LaBraMMental workload · n=36 | +8.12 pp+4.97 to +11.40 pp+8.12 pp pretraining contrast | +7.99 pp+4.74 to +11.42 pp+7.99 pp pretraining contrast |
| CBraModMental workload · n=36 | +5.90 pp+2.31 to +9.57 pp+5.90 pp pretraining contrast | +6.73 pp+3.12 to +10.43 pp+6.73 pp pretraining contrast |
Measured scores
Both readout settings stay visible
Head selection uses only inner training participants. Showing both settings prevents a test score from choosing the adaptation budget after the fact.
| Task / model | Pretrained encoder | Random encoder mean | Paired difference |
|---|---|---|---|
| LaBraMMI / rest | 53.5% | 55.9%3 initializations | −2.42 pp−5.94 to +1.03 pp |
| CBraModMI / rest | 62.9% | 56.5%3 initializations | +6.39 pp+3.06 to +9.50 pp |
| LaBraMMental workload | 64.6% | 56.5%3 initializations | +8.12 pp+4.97 to +11.40 pp |
| CBraModMental workload | 62.3% | 56.4%3 initializations | +5.90 pp+2.31 to +9.57 pp |
| Task / model | Pretrained encoder | Random encoder mean | Paired difference |
|---|---|---|---|
| LaBraMMI / rest | 52.9% | 56.5%3 initializations | −3.56 pp−7.00 to −0.19 pp |
| CBraModMI / rest | 62.7% | 54.9%3 initializations | +7.89 pp+4.53 to +11.11 pp |
| LaBraMMental workload | 64.4% | 56.5%3 initializations | +7.99 pp+4.74 to +11.42 pp |
| CBraModMental workload | 68.4% | 61.7%3 initializations | +6.73 pp+3.12 to +10.43 pp |
Fixed classical control
61.9%
61.9% balanced accuracyLog-covariance ridge · Mental workload
95% interval 56.2%–67.6%Training-seed sensitivity
A spread is not a confidence interval
Five existing EEGNet protocols were rerun with three fixed MPS seeds. The range is a small-sample training sensitivity check; it does not replace the published point estimate or describe population uncertainty.
| Protocol | Three runs | Mean | Sample SD | Full range |
|---|---|---|---|---|
| Semantic ERPn=30 · 3,600 trials | 60.50% / 60.42% / 60.33% | 60.42% | 0.08 pp | 0.17 pp60.33%–60.50% |
| P300n=21 · 2,520 trials | 55.79% / 50.36% / 52.06% | 52.74% | 2.78 pp | 5.44 pp50.36%–55.79% |
| Scalp sleep stagingn=20 · 3,000 trials | 57.33% / 54.43% / 51.60% | 54.46% | 2.87 pp | 5.73 pp51.60%–57.33% |
| BETA · 4 selected channelsn=70 · 11,200 trials | 44.77% / 44.42% / 45.03% | 44.74% | 0.30 pp | 0.61 pp44.42%–45.03% |
| BETA · 8 selected channelsn=70 · 11,200 trials | 56.54% / 56.73% / 56.99% | 56.75% | 0.23 pp | 0.46 pp56.54%–56.99% |
Dataset authors, versions, licenses and source records for all four seed-sensitivity datasets are listed under Data use & privacy → Sources in this release.
Methods & limits
What a positive contrast establishes
These controls narrow one question: whether the pretrained initialization helps this frozen architecture, split and readout relative to constructor-random encoders.
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. Both are reported.
Three seeds, one cohort
The random baseline averages three encoder initializations. Those are not three independent datasets, and 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. A method can rank only within a compatible protocol.
Exposure is unresolved
The comparison does not establish that the benchmarks were unseen during pretraining. It also does not certify a universal benefit across tasks or adaptation budgets.
MI / rest · OpenNeuro ds003810 v2.0.2
Peterson et al. · CC0-1.0.
Mental workload · EEGMAT 1.0.0
Zyma et al. · Open Data Commons Attribution License 1.0.
Data source: reviewed aggregate JSON · schema bci-report-public-deployment-topics-v1 · generated 2026-09-20.