- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2(+1)
Read review
Paper summary
The paper proposes a new method for motor-imagery-based BCI using an online baseline decoder and an offline correction (or residual) model. The authors gather data from 10 participants, with 5 sessions, each including runs where the participants moved and others where they imagined movement. For the latter, MI sessions, the target was each participant’s average velocity profile across the that movement session, separately for each target. The authors then demonstrate that this correction model generally outperforms the online model, training the decoder before each session, or training it within each session. They compare the models across 2D-screen feedback and 3D VR feedback trajectories and using correlation and RMSE.
This method could be a nice advance for noninvasive BCIs. The authors further run the experiments well, as far as I can tell, and their statistical analyses make sense (see the Strengths below). But the paper also has several weaknesses that need to be addressed for it to be considered for publication at ICLR.
Critical strengths
-
The Introduction motivates the study well.
-
The methods are well explained.
-
The authors gathered a decent cohort or a conference paper (10 participants with 10 sessions each).
-
The results are clearly in favor of the FRL.
-
The EEG collection seems serious (the system and the explanation). The Vive is not state of the art, but it seems good enough for their purposes and usage.
-
The ML procedure seems fine, as far as I can tell.
-
The authors’ Limitation section is honest and real. They should be lauded for that.
Critical weaknesses
-
The authors suggest offline analysis of the residual error is a good thing because it need not involve the user. However, they only tested their model with the stage-2 residual offline, so we do not know how well the correction 2-stage decoder works in practice—online and in real time. This is a major disadvantage for an ICLR paper.
-
To their credit, the authors mention it in the limitations, though a bit obliquely. The authors should state this more prominently in the Abstract and possibly in Fig. 1’s caption.
-
This method is further problematic because while the baseline model was trained on 100% assisted data, sessions S2-S5 were on partially assisted data. So, the test was pseudo-online replay only, also because the correction model was never used in real time.
-
In this regard, it is important to ask whether the windows that fed the offline decoder were strictly causal, not containing information from the future? It was not clear enough in the paper.
-
They tested the error model against neural data that was generated while the user was seeing feedback data that increasingly blended the decoder’s output (from the baseline model) with the target trajectory. It is unclear why the authors chose this protocol. They should motivate it and explain why it is not making their residual decoder’s job easier, especially because they do not test it online.
-
-
The lack of online results leads to another issue. Reaching movements are quite stereotyped, with the same speed profiles across trials. This is also true here (see Appendix E), especially because the target trajectories were here averaged for each participant. (It would have been different had the participants used MI to run actual BCIs and had then been tested on how well the results matched what they were attempting to achieve. Hence, the EEG from the imagined movement on every trial was compared against the same average trajectory. And the model is given information about the time steps and the decoder. If the variance is low, can we be sure that the correction model does not just memorize the average pattern (minus the decoder values) and override the decoder? This is possible given that the velocities are scaled to 0-0.3 but the decoder can scale up to values up to 1. Then, the network is trained to output values that are as close as possible to these stereotyped velocities, so it might be trained to do the above. If this interpretation is correct, a user in a closed-loop BCI setting would not be able to steer it. This could only be tested online and in real time.
-
This interpretation could explain why FRL tends to outperform the other measures in Fig. 2. Predicting the average curve might also explain why the FRL results are stable while the other measured move around much more.
-
We also see higher scores in Y and Z and less in X, which seems to be in line with the results in Appendix E, where the Y and Z curves might be more stereotypical than the X curve.
-
-
To test this, it would be good to do three things.
-
See the chosen scale (or “s”) values to see whether they are large.
-
The authors could also provide the correction model with shuffled results during training, to decorrelate it from the decoder and see what happens.
-
Finally, what happens if the residual decoder receives only time values? How much would that decrease the very high correlations in the Y direction and somewhat how ones in the Z direction?
-
-
FRL is FDG with residual training, so a fairer comparison to SAT and WSR would be to apply residual training on those two models too.
-
I found Fig. 1 more confusing than helpful. Better explanation in the caption and the text would help a lot.
-
It was hard for me to understand what the target was for the residual model. I finally figured it out, but it rook quite a bit of reading and re-reading. Also, please explain why you used decoder assistance for the feedback to the user. And why does the assistance stay at 40% of higher throughout? Has the model never gotten good enough to produce feedback on its own?
-
It is unclear, to me at least, whether this result important and non-incremental enough for the community to be featured in a top-tier conference like ICLR.
Questions for authors
The questions are included in the Weaknesses section.
4 comments
2.Response to the second weakness part
Our residual does not simply memorize an average pattern. The explanations are below:
1)We provide the mean scale values of each axis from our original optuna search:
| Feedback | Mean scale X | Mean scale Y | Mean scale Z |
|---|---|---|---|
| 2D | 0.6421 | 0.7105 | 0.6372 |
| VR | 0.5306 | 0.6150 | 0.4478 |
We found that setting the scale values too low (below 0.1) reduced the performance of the residual corrector because the correction was very small relative to it. We therefore set the scale range to 0.1–1.0. To address the reviewer’s concern that large scale values could override the decoder’s output, we also report results from retraining the residual corrector for all participants and feedback conditions with the scale values for all axes fixed at 0.1 and, separately, at 0.01. FRL with these low scale values still showed a clear advantage over FDG, although the advantage was smaller than that of the original FRL.
| Feedback | Metric | Method | X | Y | Z | Mean | p |
|---|---|---|---|---|---|---|---|
| 2D | Corr | FDG | 0.1209 ± 0.6885 | 0.7469 ± 0.3208 | 0.6551 ± 0.3678 | 0.5076 ± 0.3206 | — |
| 2D | Corr | FRL(scale = 0.1) | 0.2666 ± 0.6723 | 0.9739 ± 0.0459 | 0.8772 ± 0.2014 | 0.7059 ± 0.2436 | 0.0001 |
| 2D | RMSE | FDG | 0.0557 ± 0.0290 | 0.0581 ± 0.0229 | 0.0555 ± 0.0233 | 0.0564 ± 0.0158 | — |
| 2D | RMSE | FRL(scale = 0.1) | 0.0494 ± 0.0252 | 0.0339 ± 0.0167 | 0.0333 ± 0.0165 | 0.0389 ± 0.0099 | <0.0001 |
| VR | Corr | FDG | 0.2492 ± 0.6854 | 0.8455 ± 0.2243 | 0.8316 ± 0.2366 | 0.6420 ± 0.2870 | — |
| VR | Corr | FRL(scale = 0.1) | 0.3856 ± 0.6686 | 0.9718 ± 0.0498 | 0.9677 ± 0.0525 | 0.7751 ± 0.2264 | <0.0001 |
| VR | RMSE | FDG | 0.0533 ± 0.0318 | 0.0550 ± 0.0241 | 0.0502 ± 0.0228 | 0.0528 ± 0.0171 | — |
| VR | RMSE | FRL(scale = 0.1) | 0.0491 ± 0.0288 | 0.0330 ± 0.0177 | 0.0263 ± 0.0150 | 0.0361 ± 0.0125 | <0.0001 |
| Feedback | Metric | Method | X | Y | Z | Mean | p |
|---|---|---|---|---|---|---|---|
| 2D | Corr | FDG | 0.1209 ± 0.6885 | 0.7469 ± 0.3208 | 0.6551 ± 0.3678 | 0.5076 ± 0.3206 | — |
| 2D | Corr | FRL(scale = 0.01) | 0.1613 ± 0.6796 | 0.8022 ± 0.2636 | 0.7169 ± 0.3162 | 0.5601 ± 0.2906 | 0.0086 |
| 2D | RMSE | FDG | 0.0557 ± 0.0290 | 0.0581 ± 0.0229 | 0.0555 ± 0.0233 | 0.0564 ± 0.0158 | — |
| 2D | RMSE | FRL(scale = 0.01) | 0.0532 ± 0.0279 | 0.0530 ± 0.0225 | 0.0508 ± 0.0223 | 0.0524 ± 0.0149 | <0.0001 |
| VR | Corr | FDG | 0.2492 ± 0.6854 | 0.8455 ± 0.2243 | 0.8316 ± 0.2366 | 0.6420 ± 0.2870 | — |
| VR | Corr | FRL(scale = 0.01) | 0.2971 ± 0.6702 | 0.8743 ± 0.2001 | 0.8653 ± 0.2030 | 0.6789 ± 0.2686 | 0.0001 |
| VR | RMSE | FDG | 0.0533 ± 0.0318 | 0.0550 ± 0.0241 | 0.0502 ± 0.0228 | 0.0528 ± 0.0171 | — |
| VR | RMSE | FRL(scale = 0.01) | 0.0511 ± 0.0306 | 0.0497 ± 0.0232 | 0.0449 ± 0.0221 | 0.0486 ± 0.0161 | <0.0001 |
2)We trained the residual model on shuffled decoded–target trajectory pairs, ensuring that each decoded trajectory was paired with a reference trajectory for a different target. The model trained on shuffled pairs showed a clear drop in performance compared with the original FRL. This suggests that correct pairing contributes to FRL’s performance and provides evidence against the explanation that FRL simply memorises a single average trajectory shared across targets.
| Feedback | Metric | Method | X | Y | Z | Mean | p |
|---|---|---|---|---|---|---|---|
| 2D | Corr | FRL | 0.2865 ± 0.6264 | 0.9829 ± 0.0180 | 0.8850 ± 0.1031 | 0.7182 ± 0.2167 | 0.0015 |
| 2D | Corr | FRL(shuffled) | 0.2496 ± 0.6892 | 0.9135 ± 0.1169 | 0.8082 ± 0.2473 | 0.6571 ± 0.2629 | — |
| 2D | RMSE | FRL | 0.0497 ± 0.0255 | 0.0325 ± 0.0164 | 0.0318 ± 0.0167 | 0.0380 ± 0.0101 | 0.0004 |
| 2D | RMSE | FRL(shuffled) | 0.0501 ± 0.0256 | 0.0415 ± 0.0206 | 0.0417 ± 0.0186 | 0.0444 ± 0.0113 | — |
| VR | Corr | FRL | 0.3819 ± 0.6595 | 0.9786 ± 0.0362 | 0.9744 ± 0.0419 | 0.7783 ± 0.2218 | 0.0059 |
| VR | Corr | FRL(shuffled) | 0.3588 ± 0.6625 | 0.9153 ± 0.1232 | 0.9005 ± 0.1621 | 0.7249 ± 0.2411 | — |
| VR | RMSE | FRL | 0.0491 ± 0.0295 | 0.0316 ± 0.0178 | 0.0248 ± 0.0145 | 0.0352 ± 0.0133 | 0.0014 |
| VR | RMSE | FRL(shuffled) | 0.0527 ± 0.0297 | 0.0407 ± 0.0191 | 0.0375 ± 0.0193 | 0.0436 ± 0.0140 | — |
3)Compared with full FRL, time-only FRL showed a clear drop in performance along the Y and Z axes in both feedback conditions. In 2D, Pearson correlation decreased by 0.0643 along Y and 0.0701 along Z, with corresponding RMSE increases of 31.4% and 36.5%. In VR, Pearson correlation decreased by 0.0501 along Y and 0.0617 along Z, with corresponding RMSE increases of 26.6% and 48.8%. Please see our response to reviewer 3QZw for the full per-axis results.
The results of SAT + FRL and WSR + FRL are provided. +FRL significantly improves both SAT and WSR in every condition and metric. The hyperparameters setting follows those of FDG for each participant and feedback. Since WSR involves training and testing within the same session, we split each session’s data into an 80% training set and a 20% test set. The reported results compare mean performance of WSR and WSR+FRL on the test set.
| Feedback | Metric | Method | X | Y | Z | Mean | p |
|---|---|---|---|---|---|---|---|
| 2D | Corr | SAT | 0.1647 ± 0.6945 | 0.7894 ± 0.2719 | 0.7149 ± 0.3180 | 0.5563 ± 0.3029 | — |
| 2D | Corr | +FRL | 0.2967 ± 0.6640 | 0.9805 ± 0.0274 | 0.9045 ± 0.1567 | 0.7272 ± 0.2319 | <0.0001 |
| 2D | RMSE | SAT | 0.0596 ± 0.0318 | 0.0597 ± 0.0271 | 0.0580 ± 0.0250 | 0.0591 ± 0.0191 | — |
| 2D | RMSE | +FRL | 0.0499 ± 0.0265 | 0.0339 ± 0.0176 | 0.0315 ± 0.0172 | 0.0384 ± 0.0109 | <0.0001 |
| VR | Corr | SAT | 0.3019 ± 0.6727 | 0.8245 ± 0.2647 | 0.8090 ± 0.2689 | 0.6451 ± 0.2977 | — |
| VR | Corr | +FRL | 0.4033 ± 0.6572 | 0.9771 ± 0.0367 | 0.9614 ± 0.0716 | 0.7806 ± 0.2217 | <0.0001 |
| VR | RMSE | SAT | 0.0546 ± 0.0341 | 0.0574 ± 0.0257 | 0.0515 ± 0.0239 | 0.0545 ± 0.0197 | — |
| VR | RMSE | +FRL | 0.0474 ± 0.0301 | 0.0314 ± 0.0192 | 0.0256 ± 0.0158 | 0.0348 ± 0.0145 | <0.0001 |
| Session | Trajectory assistance | 2D correlation increase | 2D RMSE reduction | VR correlation increase | VR RMSE reduction |
|---|---|---|---|---|---|
| S2 | 65% | 42.8% | 32.9% | 18.2% | 34.1% |
| S3 | 60% | 32.9% | 37.3% | 41.0% | 42.6% |
| S4 | 50% | 18.4% | 27.1% | 12.7% | 31.9% |
| S5 | 40% | 31.3% | 40.8% | 15.8% | 34.6% |
| Feedback | Metric | Method | X | Y | Z | Mean | p |
|---|---|---|---|---|---|---|---|
| 2D | Corr | WSR | 0.3787 ± 0.6808 | 0.8255 ± 0.2528 | 0.7776 ± 0.2755 | 0.6606 ± 0.3047 | — |
| 2D | Corr | +FRL | 0.4589 ± 0.6368 | 0.9827 ± 0.0285 | 0.9342 ± 0.1142 | 0.7919 ± 0.2217 | <0.0001 |
| 2D | RMSE | WSR | 0.0500 ± 0.0355 | 0.0547 ± 0.0348 | 0.0536 ± 0.0332 | 0.0528 ± 0.0288 | — |
| 2D | RMSE | +FRL | 0.0415 ± 0.0268 | 0.0290 ± 0.0175 | 0.0266 ± 0.0164 | 0.0324 ± 0.0136 | 0.0002 |
| VR | Corr | WSR | 0.5342 ± 0.6089 | 0.8848 ± 0.1715 | 0.8740 ± 0.1837 | 0.7643 ± 0.2489 | — |
| VR | Corr | +FRL | 0.5973 ± 0.5783 | 0.9811 ± 0.0309 | 0.9711 ± 0.0668 | 0.8498 ± 0.1990 | <0.0001 |
| VR | RMSE | WSR | 0.0426 ± 0.0323 | 0.0488 ± 0.0235 | 0.0436 ± 0.0234 | 0.0450 ± 0.0195 | — |
| VR | RMSE | +FRL | 0.0369 ± 0.0286 | 0.0260 ± 0.0197 | 0.0209 ± 0.0160 | 0.0279 ± 0.0163 | <0.0001 |
| Session | Trajectory assistance | 2D correlation increase | 2D RMSE reduction | VR correlation increase | VR RMSE reduction |
|---|---|---|---|---|---|
| S1 | 100% | 21.6% | 31.9% | 9.4% | 34.4% |
| S2 | 65% | 18.9% | 37.9% | 19.9% | 42.3% |
| S3 | 60% | 10.1% | 30.4% | 7.9% | 35.7% |
| S4 | 50% | 17.7% | 41.4% | 9.7% | 37.6% |
| S5 | 40% | 33.7% | 48.2% | 9.8% | 39.1% |
The target used to train the residual model is averaged executed trajectories in real runs, as written in the first paragraph in Section 2.1 in the paper. The residual action is added to decoded output and the parameters of residual model are updated using these targets with Equations (3) and (4).
We agree to add a paragraph describing the experimental paradigm below Fig. 1 and to provide a clearer and more detailed explanation of the figure in the main text. We will upload a revised manuscript incorporating these changes. The reviewer has raised several concerns about the experimental paradigm described in our paper. The paradigm serves as a setting for evaluating our method. The focus of our paper is the residual correction method itself. The proposed corrector operates on neural decoder outputs and improves performance across all baseline decoders we tested. We additionlly reproduce another experimental protocol described in [1] using the same publicly available WAY-EEG-GAL dataset [2] and extend their protocol to evaluate our residual learning method. Adding the residual corrector provides statistically significant improvements over the baselines in all reported comparisons in this paradigm. For rEEGNet, rDeepConvNet, and rShallowConvNet, adding FRL increases mean Pearson correlation by 0.0487, 0.0265, and 0.0691, respectively, with corresponding reductions in mean RMSE of 15.9%, 7.5%, and 28.1%. Please see our second part response to Reviewer e7aZ for the detailed comparisons.
[1] Jain, A. and Kumar, L., 2025. ESI-GAL: EEG source imaging-based trajectory estimation for grasp and lift task. Computers in Biology and Medicine, 186, p.109608. [2] Luciw, M.D., Jarocka, E. and Edin, B.B., 2014. Multi-channel EEG recordings during 3,936 grasp and lift trials with varying weight and friction. Scientific data, 1(1), p.140047.
We appreciate your time and thank you for your careful reviews.
1.Response to first weakness part
Many studies on motor imagery BCIs evaluate their methods offline, including studies that use publicly available recordings. In our paper, we identify real-time evaluation of the residual corrector as future work. As stated in the paper, our work extends the experimental paradigm introduced in [1]. The authors of [1] collected the dataset, trained the CNN–LSTM decoder, and used it for online adaption of S2-S5. Given the early stage of 3D MI BCI research and, to our knowledge, the absence of another comparable ten-session dataset, this dataset provides a suitable basis for establishing our proposed FRL approach. We use their dataset and adapt their experimental paradigm to conduct a full offline evaluation of our proposed residual correction method in a pseudo-online setting across eight unseen sessions, a standard step before proceeding to online evaluation.
In the CNN–LSTM implementation, the decoder uses backward-looking input windows. For a prediction associated with time t, the model uses ERSP features from the current window and from earlier windows, all of which end at or before t. Temporal resizing is performed within each selected window. EEG filtering uses only current and past samples, and spectral features are computed from windows ending at their corresponding time points. So all decoding within a trial is causal and all test session are tested causally aligning to 8 different sessions.
Trajectory assistance is used to provide a degree of stability, gradually ease participants into learning to control the visual feedback and prevent erratic movement. The decoder used in S2–S5 does not consistently generate accurate trajectories toward the targets, as it does not achieve a high correlation with the target trajectory. Trajectory assistance therefore provides participants with positively biased feedback toward the targets.
The residual corrector only receives the decoder predictions and temporal context as input. Target trajectories are used only as training labels, and the residual corrector is evaluated on test sessions. Trajectory assistance does not directly provide target information to the residual corrector during inference. Moreover, to address the reviewer’s concern that trajectory assistance may facilitate residual correction, we hypothesised that higher assistance levels would be associated with larger FRL gains over FDG. We examined this prediction across the held-out sessions S2–S5. The table reports the percentage increase in mean Pearson correlation and the percentage reduction in RMSE relative to FDG (Figs. 2–3 in our paper).
We have also provided the SAT+FRL and WSR + FRL results across sessions (See responses in the next part). Because assistance level is tied to session, these comparisons are descriptive. Across the tested sessions/trials, the results provide no consistent evidence that stronger trajectory assistance was associated with larger FRL gains.
[1] McShane, N., Korik, A., McCreadie, K., Bois, N.D., Charles, D. and Coyle, D., 2026. Embodied Virtual Reality Feedback Reshapes Neural Representations to Support Continuous Three-Dimensional Motor Imagery Decoding. arXiv preprint arXiv:2605.29677.