- Rating
- 1
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 3
- Soundness
- 1
Read review
Paper summary
This paper trains an AE to do forward/inverse solving. In training AE, the latent is required to match teacher.
Critical strengths
This paper is easy to follow. The demonstration is clear. The theorems are intuitive and easy to understand.
Critical weaknesses
The theorem part is trivial. The main contribution of this paper shoud lie in the empirical part. However, the experiements are not convincing. See questions.
Questions for authors
- FO (without teacher) is also a mask deterministic AE. Why it already largely surpasses all mask AE baselines? It there any techniques inside mask AE? Why exactly the same formulation differs so much in performance?
- The training budgets and model capacity of SCOPE is much larger than the ones of baselines, this makes comparision not fair.
- In the ablation, it seems that teacher matching (Table. 7) brings suble improvement, and it is even worse than AE in certain cases. Does it mean teacher matching is not very useful?
7 comments
Question 1
FO (without teacher) is also a mask deterministic AE. Why it already largely surpasses all mask AE baselines? It there any techniques inside mask AE? Why exactly the same formulation differs so much in performance?
Response:
Thank you for this perceptive question about the teacher-free model. We appreciate the need to explain why models with a common reconstruction formulation can perform differently. The original models share the masked-reconstruction task, while their architectures, capacities and training configurations differ. We clarify those differences below.
[D1] What FO contains.
For clarity, FO uses the SCOPE inference architecture (Appendix E.5): a patch-token Transformer encoder that receives the observed values together with their masks, a mask conditioner that encodes how much of each patch is observed, a latent predictor and the native decoder, totaling 40,040,834 parameters. We train it from random initialization using only the physical field-reconstruction loss, without complete-view pretraining or a teacher.
We use the new approximately 40M comparison to discuss external performance at a more closely aligned scale.
[D2] Evidence from the expanded comparison.
To help contextualize FO's performance, we compare it with the approximately 40M FNO and Transolver baselines. Their forward/inverse results are shown below, with the FO values taken from Table 7:
| PDE | FO | FNO40M | Transolver40M |
|---|---|---|---|
| Poisson | 1.832/11.246 | 3.205/15.097 | 2.605/12.718 |
| Helmholtz | 1.831/10.707 | 3.228/14.817 | 2.244/12.698 |
| Darcy (inverse: BER) | 2.163/1.583 | 2.626/3.594 | 2.444/2.641 |
| NS | 2.627/7.329 | 3.975/11.162 | 6.072/13.771 |
| NS (BCs) | 1.983/0.708 | 2.317/0.736 | 2.615/1.156 |
We report the values as percentages. FO uses no technique beyond the architecture and field supervision described in [D1].
To make the interpretation clearer, the revised manuscript reports the networks, inputs and training budgets of the compared models (Tables 2, 18, 20 and 21).
Manuscript changes. Following your helpful question, the revised manuscript states FO's architecture, initialization and loss in the footnote of Table 1 (page 8, lines 410–412) and in the “Field-only comparison” paragraph of Section 4.3 (page 9, lines 456–457), and lists FO as a same-architecture recipe reference beside the approximately 40M baselines in Table 22 (Appendix K.1, page 36). The approximately 40M results serve as the external comparison, and latent matching is discussed separately through Full–FGV (Section 4.3, page 9, lines 460–463, and Appendix K.2).
Question 2
The training budgets and model capacity of SCOPE is much larger than ...
Response:
We sincerely appreciate your concern about the fairness of the capacity and resource comparison. In response, we have expanded the comparison to four approximately 40M baselines—FNO, Transolver, CNO and DeepONet—evaluated against SCOPE Full-native across all five PDEs and both recovery directions. We summarize these results and the resource accounting below.
[D1] Completed approximately 40M comparison.
To keep the evaluation protocol consistent, all methods use the same test records, normalization and observation masks. Each entry reports forward/inverse error in percent, with BER for Darcy inverse and the whole-grid metric for NS (BCs).
| Model | Parameters (M) | Poisson | Helmholtz | Darcy | NS | NS (BCs) |
|---|---|---|---|---|---|---|
| SCOPE Full-native | 40.04 | 1.680/9.502 | 1.636/8.702 | 2.174/1.556 | 2.581/6.973 | 2.115/0.649 |
| FNO40M | 39.40 | 3.205/15.097 | 3.228/14.817 | 2.626/3.594 | 3.975/11.162 | 2.317/0.736 |
| Transolver40M | 40.19 | 2.605/12.718 | 2.244/12.698 | 2.444/2.641 | 6.072/13.771 | 2.615/1.156 |
| CNO40M | 39.92 | 11.314/24.201 | 13.810/25.253 | 15.084/2.955 | 12.104/17.254 | 14.616/7.385 |
| DeepONet40M | 40.05 | 10.375/27.765 | 9.949/35.280 | 5.597/8.012 | 21.121/32.437 | 26.614/22.046 |
[D2] Conclusions supported by the expanded comparison.
Under the evaluated training configurations, we observe lower error for Full-native than for FNO40M, Transolver40M, CNO40M and DeepONet40M on all ten forward/inverse endpoints.
For the inverse tasks on Poisson, Helmholtz and NS, we obtain 9.502%, 8.702% and 6.973%, respectively, compared with 12.718%, 12.698% and 11.162% for the strongest approximately 40M baseline on the corresponding endpoints.
This ordering is stable. The table below reports the K=500 errors as mean ± standard deviation over five test-mask seeds:
| Endpoint | Full | FNO40M | Transolver40M | CNO40M | DeepONet40M |
|---|---|---|---|---|---|
| Poisson F | 1.648 ± 0.019 | 3.137 ± 0.047 | 2.585 ± 0.020 | 11.312 ± 0.031 | 10.368 ± 0.008 |
| Poisson I | 9.510 ± 0.010 | 15.102 ± 0.011 | 12.724 ± 0.011 | 24.215 ± 0.014 | 27.805 ± 0.032 |
| Helmholtz F | 1.639 ± 0.010 | 3.229 ± 0.012 | 2.244 ± 0.009 | 13.822 ± 0.009 | 9.954 ± 0.007 |
| Helmholtz I | 8.706 ± 0.002 | 14.810 ± 0.006 | 12.695 ± 0.005 | 25.253 ± 0.001 | 35.275 ± 0.004 |
| Darcy F | 2.172 ± 0.005 | 2.614 ± 0.009 | 2.441 ± 0.006 | 15.082 ± 0.002 | 5.593 ± 0.002 |
| Darcy I (BER) | 1.558 ± 0.001 | 3.594 ± 0.001 | 2.643 ± 0.002 | 2.955 ± 0.001 | 8.014 ± 0.002 |
| NS F | 2.576 ± 0.008 | 3.960 ± 0.011 | 6.065 ± 0.013 | 12.106 ± 0.013 | 21.110 ± 0.009 |
| NS I | 6.979 ± 0.004 | 11.171 ± 0.008 | 13.784 ± 0.017 | 17.264 ± 0.008 | 32.434 ± 0.002 |
| NS (BCs) F | 2.094 ± 0.018 | 2.290 ± 0.015 | 2.656 ± 0.033 | 14.607 ± 0.033 | 26.788 ± 0.119 |
| NS (BCs) I | 0.649 ± 0.000 | 0.740 ± 0.005 | 1.151 ± 0.003 | 7.351 ± 0.025 | 22.052 ± 0.010 |
Full has the lowest mean error on all ten endpoints, and its standard deviation across mask seeds is at most 0.019. In addition, the paired per-record bootstrap intervals of all 40 Full-minus-baseline differences lie below zero (table in our response to Reviewer MiBT, M-W2 / M-Q2).
To keep the comparison balanced, we retain all task-specific results in the complete table.
[D3] Computational-resource accounting.
Following your emphasis on computational resources, we also compiled per-method training-resource records covering main-training GPU-hours and hardware. We report parameter-scale alignment and resource accounting separately, so the accuracy measurements can be interpreted under their actual training configurations.
Full's main training takes 45.0, 41.7, 36.3, 28.6 and 12.7 GPU-hours on Poisson, Helmholtz, Darcy, NS and NS (BCs). Transolver40M is trained for 41.4, 34.2, 43.6, 41.6 and 22.1 GPU-hours, CNO40M for 27.8, 44.9, 58.1, 52.1 and 5.7 GPU-hours, and FNO40M for 10.3, 10.1, 10.1, 10.4 and 2.8 GPU-hours. On Darcy and NS, Transolver40M and CNO40M are trained longer than Full, and Full has lower error on all four endpoints of these PDEs. The training hours and training-loss trajectories of Full and Transolver40M are given in our response to Reviewer MiBT (M-W2 / M-Q2). The training configurations and the main-training GPU-hours of every method are given in Tables 20 and 21 of the revised manuscript (Appendix K.1, page 35).
Manuscript changes. In response to your feedback, the revised manuscript makes the complete approximately 40M comparison central to the evaluation (Table 1 and Section 4.2, page 8, lines 414–416) and provides its training configurations and resource accounting in a new Appendix K.1 (page 35, Tables 20 and 21). The ten-endpoint comparison with the four baseline families is reported with its paired intervals and test-mask seeds in Tables 22–24 (page 36), with the evaluation conditions made clear.
Weakness 1
Response:
We sincerely thank you for this candid and helpful assessment. We appreciate your emphasis on the empirical evidence. In response, we have completed the approximately 40M comparison to strengthen the empirical evaluation. Below, we summarize that evidence and clarify the explanatory role of the analysis alongside the requested experiments.
[D1] Expanded empirical evidence.
The expanded evaluation includes FNO40M, Transolver40M, CNO40M and DeepONet40M on all five PDEs in both directions. Under the evaluated configurations, we observe lower error for Full-native than for FNO, Transolver, CNO and DeepONet on all ten endpoints.
For ease of reference, we provide the complete table and computational-resource interpretation in our response to Q2. The pretrained FGV and native-size decoder refits provide separate controls for the component-attribution questions raised in the reviews.
[D2] What the theory is intended to establish.
We would like to clarify the intended role of the fixed-reference squared-risk analysis. It distinguishes observation uncertainty, decoder incompatibility and predictor excess risk to explain why optimal latent prediction need not yield optimal physical reconstruction. Consistent with this, full-target prediction without main-stage grounding (FJV) has higher error than Full on all ten endpoints of Table 7, for example 11.6476% versus 8.7023% on Helmholtz inverse.
The weighting analysis examines the target metric under its stated reference assumptions. In the corresponding experiment, uniform weighting gives 9.5024% Poisson inverse error and a mask-dependent equal-group weighting gives 10.7760% (our response to Reviewer VrNy, V-W2).
The readout-transfer result gives sufficient conditions connecting complete-view and sparse-view decoder behavior. In Table 10, a decoder fitted on complete-view representations (DP-10M) lowers the NS (BCs) inverse error from 0.6491% to 0.4576% but not the Darcy inverse BER (1.5564% to 1.6468%), whereas a decoder fitted on the sparse path (SP-10M) lowers both, to 0.4524% and 1.5503%.
The contribution of these results lies in their relevance to the design, and the observed benefits of the full training procedure are assessed through the empirical controls.
Manuscript changes. Guided by your feedback, the revised manuscript foregrounds the full approximately 40M comparison (Table 1 and Section 4.2, page 8, lines 414–416). Section 3.3 now opens by stating the three questions examined by the analysis and the experiment that addresses each of them (page 6, lines 301–303), which connects each empirical conclusion to its corresponding control and makes explicit that the measured performance is supported by experiments.