- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper presents GeoLPE, an EEG encoder combining continuous absolute electrode coordinates, relative geometric attention, and structured latent prediction with an EMA teacher. The pretraining objective also includes variance, waveform-reconstruction, and boundary terms. Evaluation covers eight EEG datasets under leave-one-subject-out and within-subject protocols, using both fine-tuning and frozen linear probing. A separate adaptation architecture combines the pretrained low-frequency encoder with a high-gamma branch for ECoG finger-trajectory prediction.
Critical strengths
-
Consistent transfer improvements under two adaptation regimes. In Table 1, GeoLPE improves the cross-dataset LOSO mean over REVE from 67.54% to 69.70% with fine-tuning and from 60.25% to 62.58% with linear probing. The frozen-encoder improvement is particularly useful evidence that the representation itself carries transferable information. The within-subject results also improve on seven of eight datasets with fine-tuning and all eight with linear probing (Table 2).
-
Substantial component analysis with appropriately qualified representation analyses. Table 4 compares waveform and latent targets, masking strategies, auxiliary losses, and geometric attention components. The held-out prototype analyses fit their transforms and class prototypes on the reference/training partition (Appendix 13.4–13.5), which makes them more informative than a visualization alone. The paper also explicitly limits its ECoG claim to later segments from seen participants and identifies published baseline numbers as non-unified comparisons.
Critical weaknesses
-
The central geometric-transfer claim is not directly tested, and the mechanism remains only partially isolated. The eight datasets span different montages, but ordinary within-dataset evaluation does not establish robustness when the electrode layout changes between adaptation and evaluation. Table 4 shows that geometry-related components help the aggregate score; it does not show whether this benefit comes from handling layout shifts, additional capacity, or a useful inductive bias on fixed layouts. This distinction matters because montage variability is a central motivation. A focused experiment with held-out electrode subsets or layouts, comparing the full model with the existing geometry ablations, would address the claim directly. Coordinate perturbation or a shuffled-coordinate control would further distinguish geometric information from the parameterization. Per-dataset, paired uncertainty for these comparisons would also help: the aggregate ablation scores currently obscure the consistency of the effects. A matched latent-prediction comparator would strengthen novelty attribution, but a large new benchmark suite is not necessary.
-
The ECoG experiment does not isolate the value of EEG pretraining. The reported 0.566 mean correlation is produced by a substantial adaptation system with a high-gamma branch, cross-frequency training, and three retained prediction branches (Section 3 and Appendix 11). The comparison with the published CORTEG value of 0.554 cannot establish either a reliable advantage or that the pretrained EEG encoder is responsible. The manuscript acknowledges the first limitation, but the second remains important to its cross-modality contribution. The decisive control is the same complete adaptation architecture with random initialization of the EEG encoder; a high-gamma-only control would identify how much signal comes from the added branch. Report paired participant-level differences under identical chronological splits. If these controls are unavailable, the ECoG result should be framed as feasibility of the complete adaptation system rather than evidence that EEG pretraining improves ECoG decoding.
Questions for authors
-
Geometry clarification and targeted validation: Can the authors provide a controlled change-of-montage evaluation using the full model and the existing geometry ablations? Since the distance bias normalizes by the maximum distance in the current montage, how sensitive is the encoder to removing electrodes that change this normalization?
-
Attribution: What is the ECoG performance of the identical three-branch adaptation system with a randomly initialized low-frequency encoder, and of a high-gamma-only variant? These results could materially change my assessment of the cross-modality contribution.
-
Reproducibility: Can the authors provide the EEG fine-tuning/linear-probing configurations, and an executable pretraining manifest with sample counts and sampling weights? The 97 entries are explicitly not 97 independent cohorts; how are duplicate recordings and aliases across sources checked when excluding downstream data?
4 comments
Response to Reviewer 3 (Part 2/4)
| dataset | Training layout A | Test layout B |
|---|---|---|
| BCIC-IV-2a | FC1、FCZ、FC2、C5、C3、CZ、C4、C6、CP1、CP2、P1、P2 | FZ、FC3、FC4、C3、C1、CZ、C2、C4、CP3、CPZ、CP4、PZ、POZ |
| BCIC-IV-2b | C4 CZ | C3 CZ |
| BNCI2015001 | FC3、FC4、C5、C3、CZ、C4、C6、CPZ | FCZ、C3、C1、CZ、C2、C4、CP3、CP4 |
| BNCI2014008 | FZ、CZ、PZ、PO7、PO8 | CZ、PZ、OZ、P3、P4 |
| BNCI2014009 | FZ、CZ、PZ、PO7、PO8、FCZ、C3、C4、CP3、CP4 | CZ、PZ、OZ、P3、P4、F3、F4、C3、C4、CPZ |
| CastillosCVEP100 | FZ、F3、FC1、C3、T7、CP1、P7、O1、OZ、O2、P8、CP2、T8、C4、FC2、F4 | FP1、F7、FC5、C3、CP5、PZ、P3、O1、OZ、O2、P4、CP6、C4、CZ、FC6、F8、FP2 |
| Hinss2021 | FP1、FZ、F3、FT9、C3、T7、CP5、CP1、PZ、P3、O1、O2、P4、CP6、CP2、FCZ、C4、T8、F4、FP2、AFZ、F5、FC3、C1、P5、PO7、PO8、P6、C2、FC4、FT10、F6 | FZ、F7、FC5、FC1、C3、PZ、P7、OZ、P8、C4、FT8、FC6、FC2、F8、AF7、AF3、F1、FT7、C5、TP7、CP3、P1、PO3、POZ、PO4、P2、CPZ、CP4、TP8、C6、AF8、AF4、F2 |
| Kojima2024A | AF3、AFZ、AF4、F5、F1、FZ、F2、F6、FT9、FC5、FC3、FCZ、FC4、FC6、FT10、C3、CZ、C4、TP9、CP5、CP3、CPZ、CP4、CP6、TP10、P7、P3、PZ、P4、P8、PO7、PO8、O1、O2 | FP1、FP2、AF7、AF8、F7、F3、FZ、F4、F8、FT7、FC1、FC2、FT8、T7、C5、C3、C1、CZ、C2、C4、C6、T8、TP7、CP1、CP2、TP8、P5、P1、PZ、P2、P6、PO3、POZ、PO4、OZ |
Cross-layout evaluation
| Dataset (participants) | GeoLPE | REVE | CBraMod | LaBraM |
|---|---|---|---|---|
| BCIC-IV-2a(9) | 42.51 | 35.56 | 35.23 | 38.61 |
| BCIC-IV-2b (9) | 59.67 | 53.26 | 56.72 | 52.28 |
| BNCI2015001(12) | 62.44 | 58.22 | 48.84 | 59.23 |
| BNCI2014008(8) | 61.35 | 59.04 | 57.84 | 59.31 |
| BNCI2014009(10) | 71.59 | 68.67 | 59.07 | 68.67 |
| CastillosCVEP100(12) | 76.44 | 75.97 | 66.46 | 74.46 |
| Hinss2021(15) | 51.30 | 48.52 | 44.68 | 42.30 |
| Kojima2024A(11) | 65.71 | 63.07 | 62.39 | 60.15 |
| Macro BAcc (%) | 61.38 | 57.79 | 53.90 | 56.88 |
Geometry ablations under the same A→B protocol
We evaluated the existing geometry ablations under this same cross-layout setting. Fixed-frequency variants keep the scalar spatial-frequency coefficients at their deterministic initialization, , and fix the mixed-coordinate direction parameters to zero from the start of pretraining. Identity-basis variants set throughout training.
| Spatial RoPE | Cayley basis | learnable frequencies | geometry bias | Original LOSO full FT | Within-layout A | Cross-layout A→B |
|---|---|---|---|---|---|---|
| w/ spatial RoPE | ✓ | ✓ | 68.85 | 62.48 | 60.36 | |
| w/ spatial RoPE | ✓ | ✓ | 69.24 | 64.98 | 61.17 | |
| w/ spatial RoPE | ✓ | ✓ | 67.87 | 63.46 | 60.58 | |
| w/ spatial RoPE | ✓ | 67.98 | 61.55 | 59.92 | ||
| w/ spatial RoPE | ✓ | 67.20 | 61.89 | 59.52 | ||
| w/ spatial RoPE | ✓ | 67.03 | 62.76 | 59.13 | ||
| w/ spatial RoPE | ✓ | ✓ | ✓ | 69.70 | 64.23 | 61.38 |
| w/ spatial RoPE | 66.53 | 63.29 | 59.18 | |||
| w/o spatial RoPE | ✓ | 66.19 | 60.95 | 58.74 | ||
| w/o spatial RoPE | 64.16 | 58.30 | 55.32 |
To examine the consistency of the aggregate effects across datasets, we additionally report per-dataset A→B results for the full model and two key geometry ablations below. The w/o spatial RoPE variant retains the geometric attention bias, while the w/o geometry bias variant retains the full spatial RoPE mechanism.
| dataset | Full BAcc | w/o spatial RoPE BAcc | w/o geometry bias BAcc |
|---|---|---|---|
| BCIC-IV-2a | 42.51 | 39.20 ± 0.12 | 41.70 ± 0.11 |
| BCIC-IV-2b | 59.67 | 57.63 ± 0.23 | 58.94 ± 0.06 |
| BNCI2015001 | 62.44 | 60.28 ± 0.14 | 62.32 ± 0.09 |
| BNCI2014008 | 61.35 | 59.39 ± 0.18 | 60.15 ± 0.14 |
| BNCI2014009 | 71.59 | 68.73 ± 0.16 | 71.39 ± 0.10 |
| CastillosCVEP100 | 76.44 | 72.57 ± 0.12 | 75.21 ± 0.14 |
| Hinss2021 | 51.30 | 48.62 ± 0.15 | 49.38 ± 0.06 |
| Kojima2024A | 65.71 | 63.50 ±0.06 | 65.55 ± 0.10 |
| Macro BAcc(%) | 61.38 | 58.74 | 60.58 |
Full GeoLPE achieves macro BAcc of 64.23% under A→A and 61.38% under A→B. Under A→B, the full model outperforms both ablations on all eight datasets, showing a consistent direction of the observed effects across datasets. It attains the highest A→B score among the evaluated geometry variants, although it does not uniformly exhibit the smallest decrease between layouts. The identity-basis variant performs better under A→A (64.98%) but slightly worse under A→B (61.17%), suggesting a benefit of the learned basis under the tested layout changes. These results support the utility of the geometric components for cross-layout performance, while showing that their effects on within-layout performance and sensitivity to layout changes differ.
The original-layout LOSO component ablations have been added to the Experiments section and the appendix.
Response to Reviewer 3 (Part 3/4)
We additionally evaluated GeoLPE’s sensitivity to coordinate errors at inference using the original LOSO full fine-tuning models and full electrode layouts. The EEG signals and all model parameters were kept fixed. We independently perturbed each electrode’s three-dimensional coordinate as , where and mm denotes the standard deviation along each coordinate axis.
Test-time coordinate perturbations
| Dataset (subjects) | Original coordinates | 2.5 mm | 5 mm | 10 mm |
|---|---|---|---|---|
| BCIC-IV-2a (9) | 56.13 ± 0.96 | 56.05 | 55.75 | 55.38 |
| BCIC-IV-2b (9) | 76.38 ± 0.58 | 76.12 | 75.72 | 74.27 |
| BNCI2015001 (12) | 66.08 ± 0.62 | 65.97 | 65.64 | 63.25 |
| BNCI2014008 (8) | 69.35 ± 0.65 | 69.29 | 68.96 | 67.65 |
| BNCI2014009 (10) | 78.72 ± 0.73 | 78.58 | 78.00 | 77.78 |
| CastillosCVEP100 (12) | 88.19 ± 0.61 | 88.09 | 87.76 | 87.19 |
| Hinss2021 (15) | 53.28 ± 1.21 | 53.02 | 52.55 | 51.43 |
| Kojima2024A (11) | 69.44 ± 0.92 | 69.39 | 68.92 | 68.13 |
| Macro Bacc(%) | 69.70 | 69.56 | 69.16 | 68.14 |
Sensitivity to distance normalization
To isolate the normalization effect, we evaluated the same layout-B recordings with the same A-adapted model, varying only the denominator of the normalized distance term. We used either the maximum pairwise electrode distance in the source layout, (), or that in the current target layout, ().
| Normalization denominator at inference | Macro BAcc (%) |
|---|---|
| Source-layout maximum distance () | 61.20 |
| Current target-layout maximum distance () | 61.38 |
The difference is 0.18 percentage points, indicating limited aggregate sensitivity to this normalization choice under the tested layout changes. These controlled evaluations directly address the reviewer's concerns regarding montage changes, geometric sensitivity, and distance normalization, complementing the original downstream adaptation experiments.
Weakness 2 and Question 2: Contribution of EEG pretraining to ECoG decoding
Weakness 2
The ECoG experiment does not isolate the value of EEG pretraining. The reported 0.566 mean correlation is produced by a substantial adaptation system with a high-gamma branch, cross-frequency training, and three retained prediction branches (Section 3 and Appendix 11). The comparison with the published CORTEG value of 0.554 cannot establish either a reliable advantage or that the pretrained EEG encoder is responsible. The manuscript acknowledges the first limitation, but the second remains important to its cross-modality contribution. The decisive control is the same complete adaptation architecture with random initialization of the EEG encoder; a high-gamma-only control would identify how much signal comes from the added branch. Report paired participant-level differences under identical chronological splits. If these controls are unavailable, the ECoG result should be framed as feasibility of the complete adaptation system rather than evidence that EEG pretraining improves ECoG decoding.
Question 2
Attribution: What is the ECoG performance of the identical three-branch adaptation system with a randomly initialized low-frequency encoder, and of a high-gamma-only variant? These results could materially change my assessment of the cross-modality contribution.
Response to W2 and Q2
Within the same complete adaptation architecture, EEG pretraining of the low-frequency encoder increases mean Pearson from 0.5340 to 0.5660, an absolute gain of 0.0320, with positive paired differences for all nine participants. This matched comparison supports an additional benefit from EEG pretraining within the evaluated adaptation system. The HGA-only control achieves a mean correlation of 0.4539, below both full-system variants, showing that the complete system achieves better decoding performance than the high-gamma pathway alone. For the pretrained-versus-random comparison, both configurations retain the high-gamma branch, cross-frequency training, and all three prediction branches, and use identical downstream training settings and chronological splits. The participant-level Pearson correlations are reported below.
| Participant | Pretrained full system | Randomly initialized full system | HGA-only |
|---|---|---|---|
| bp | 0.5817 | 0.5378 | 0.4394 |
| cc | 0.674 | 0.6726 | 0.6201 |
| ht | 0.4597 | 0.4182 | 0.3188 |
| jc | 0.6063 | 0.5398 | 0.4246 |
| jp | 0.6017 | 0.5901 | 0.5038 |
| mv | 0.7167 | 0.6948 | 0.6072 |
| wc | 0.4970 | 0.4545 | 0.3928 |
| wm | 0.311 | 0.2632 | 0.2078 |
| zt | 0.6459 | 0.6347 | 0.5702 |
| Mean | 0.5660 | 0.534 | 0.4539 |
We have added these matched ECoG controls to the Experiments section, with participant-level results and control settings in the appendix.
Response to Reviewer 3 (Part 4/4)
Reproducibility: Adaptation settings and pretraining provenance
Reproducibility: Can the authors provide the EEG fine-tuning/linear-probing configurations, and an executable pretraining manifest with sample counts and sampling weights? The 97 entries are explicitly not 97 independent cohorts; how are duplicate recordings and aliases across sources checked when excluding downstream data?
Response
We provide additional details below regarding the downstream adaptation settings, pretraining sampling procedure, and dataset provenance.
The configurations used for the within-subject and LOSO EEG evaluations reported in the paper are:
| Configuration | Full fine-tuning | Linear probing |
|---|---|---|
| Optimizer | ||
| Learning rate | ||
| Weight decay | 0.01 | 0.001 |
| Batch size | 64 | 64 |
| Training epochs | 50 | 50 |
| Warmup | None | Exponential warmup over the first 10 epochs, applied by training step |
| Classification head | Classification-query pooled feature together with the flattened final-layer features, followed by RMSNorm and a single linear layer, following the REVE readout design.[1] | Same |
| Trainable parameters | Encoder and classification head | classification head; encoder frozen |
Pretraining manifest and sampling weights. The complete source-to-citation mapping for the 97 pretraining entries is already provided in Appendix 7.1, with the dataset provenance summary in Table 7. We will publicly release an executable pretraining manifest containing source identifiers, per-entry preprocessed segment counts, and sampling probabilities, together with the data-preparation code, to enable independent verification of the corpus composition and reproduction of the pretraining procedure. For sampling, we follow REVE[1]: we sample uniformly over the pooled set of preprocessed EEG segments.
Duplicate recordings and dataset aliases. We track dataset provenance through the original publication and data-release source associated with each entry. We check dataset aliases against this provenance mapping so that entries referring to the same original dataset or release are identified consistently even when different loader names or task-specific names are used.
References
[1] El Ouahidi Y, et al. REVE: A foundation model for EEG-adapting to any setup with large-scale pretraining on 25,000 subjects
Response to Reviewer 3 (Part 1/4)
We thank the reviewer for the detailed assessment and suggestions for targeted controls. Below, we report controlled cross-layout evaluations and geometry ablations, coordinate-sensitivity analyses, matched ECoG controls, and additional reproducibility details.
Weakness 1 and Question 1: Controlled cross-layout evaluation
Weakness 1
Question 1
Response to W1 and Q1
In our added cross-layout evaluation across all eight downstream datasets, GeoLPE achieves the highest BAcc on every dataset, with a macro BAcc of 61.38%, compared with 57.79% for the strongest baseline, REVE. Under the same protocol, removing spatial RoPE while retaining the geometric attention bias reduces macro BAcc by 2.64 percentage points, while removing the bias with spatial RoPE retained reduces it by 0.80 percentage points. The full model achieves macro BAcc of 64.23% under A→A and 61.38% under A→B. It achieves the highest A→B score among the evaluated geometry variants, although it does not uniformly exhibit the smallest decrease between layouts. These results support the utility of the geometric components for cross-layout performance.
We would like to clarify that “geometric-transfer” was not the central claim of our original manuscript. Our geometry-related contribution is the joint encoding of absolute spatiotemporal positions and relative inter-electrode geometry in a unified continuous coordinate system, supporting heterogeneous electrode montages during multi-dataset pretraining and downstream adaptation. Our original within-subject and LOSO evaluations across eight datasets assess pretrained representations through dataset-specific downstream adaptation, using both full fine-tuning and linear probing. Directly applying an encoder and readout adapted on one montage to another montage with all parameters fixed is a distinct evaluation setting. To address the reviewer’s question, we additionally examine this setting through controlled cross-layout evaluation and geometry ablations, and separately assess sensitivity to coordinate errors.
In a separate evaluation using the original LOSO full fine-tuning models and electrode layouts, macro BAcc decreases from 69.70% to 68.14% under test-time Gaussian coordinate perturbations with a per-axis standard deviation of 10 mm, the largest tested noise scale. This indicates limited aggregate sensitivity to the tested local coordinate errors.
For the paired A→A and A→B evaluations only, to accommodate the layout change, GeoLPE, all baselines, and the geometry ablations used mean pooling over valid electrode and temporal positions in the final encoder layer, followed by LayerNorm and a linear classifier, yielding a fixed-dimensional representation regardless of electrode count. Each encoder and readout was adapted using layout A and then evaluated on the same held-out recordings under both layout A (A→A) and layout B (A→B), with all parameters fixed during both evaluations. The same data splits and layout pairs were used for all methods and ablations. The layouts partially overlap, and layout B includes electrodes absent during adaptation, testing changes in electrode coverage while holding the dataset and task fixed. Macro BAcc is the unweighted average across the eight datasets.