- Rating
- 3(+2)
- Confidence
- 4
- Informativeness
- 1
- Presentation
- 2(+1)
- Soundness
- 2(+1)
Read review
Paper summary
This paper proposes LGO, an additional module designed to be integrated into existing neural operators to provide geometric and boundary condition information for conditioning the models. For each query point, the module extracts boundary points by type and computes relative geometric information to learn a representation.
Critical strengths
- In the work, the authors identify the importance of conditioning neural operators using the boundary conditions of the simulations as an important source of information.
Critical weaknesses
-
Authors claim that "Neural operators efficiently model global flow fields but typically lack explicit local boundary information at each query point.". However, neural operators usually receive boundary information as input in the form of mask, representing, for instance, fluid node, wall node, source node, etc. That information passing through the inductive bias mechanism of the model is usually enough to represent the boundary information of the problem. This has been represented in several previous works, for example, MeshGraphNets [1]. Authors do not clarify if they included this information as input to their "Host" models, and if so, in which particular way they did it.
-
I consider the baselines and experiments insufficient. There are several stronger baselines that, given the correct input information (e.g. boundary condition information, mesh info, etc) could potentially be stronger than the ones used in the paper. To name a few: Transolver++ [2], AB-UPT [3], GeoTransolver [4], in addition to the Transolver model the authors mention in the paper, which you could find an easy to use implementation in PhysicsNeMo library. I suspect that the attention-based inductive bias of these models, in addition to correct input information, could potentially model small scales with good performance.
[1] Pfaff, Tobias, et al. "Learning mesh-based simulation with graph networks." arXiv preprint arXiv:2010.03409 (2020).
[2] Luo, Huakun, et al. "Transolver++: An accurate neural solver for pdes on million-scale geometries." arXiv preprint arXiv:2502.02414 (2025).
[3] Alkin, Benedikt, et al. "AB-UPT: Scaling neural CFD surrogates for high-fidelity automotive aerodynamics simulations via anchored-branched universal physics transformers." arXiv preprint arXiv:2502.09692 (2025).
[4] Adams, Corey, et al. "Geotransolver: Learning physics on irregular domains using multi-scale geometry aware physics attention transformer." arXiv preprint arXiv:2512.20399 (2025).
Questions for authors
In addition to the weaknesses mentioned above, please clarify:
- What is the reason you decided to use R^2 as the primary metric for baseline comparison, rather than other more conventional metrics in the surrogate modeling field such as L2 error? I think you could benefit from including additional metrics beyond R^2.
5 comments
Response to Reviewer dJhB (2/4)
R3-W2: "I consider the baselines and experiments insufficient. There are several stronger baselines that, given the correct input information (e.g. boundary condition information, mesh info, etc) could potentially be stronger than the ones used in the paper. To name a few: Transolver++ [2], AB-UPT [3], GeoTransolver [4], in addition to the Transolver model the authors mention in the paper [...]. I suspect that the attention-based inductive bias of these models, in addition to correct input information, could potentially model small scales with good performance."
We added GeoTransolver and Transolver++ as baselines, using the PhysicsNeMo code unchanged. Each gets the same 13-channel boundary input as every other model, the same search over four learning rates (1e-3 was best for both), 450 epochs, and one scoring pass on test. GeoTransolver's local features are off by default in PhysicsNeMo; we switched them on, with ball radii of 0.05 and 0.15 of the normalised room and 8 and 32 neighbours, as the Table 3 caption now states. We did not find a setting that brings it level with Transolver here; at the two are within 0.03. Held-out test, single seed:
| Model | Parameters | Full-field , (near jet) | Full-field , (near jet) | Relative , / | Time per full field |
|---|---|---|---|---|---|
| Transolver (already in the paper) | 5.99 M | 0.967 (0.992) | 0.582 (0.584) | 0.179 / 0.615 | 2.6 s |
| GeoTransolver | 5.96 M | 0.940 (0.976) | 0.524 (0.513) | 0.243 / 0.665 | 1.3 s |
| Transolver++ | 5.47 M | 0.961 (0.987) | 0.542 (0.539) | 0.195 / 0.652 | 3.1 s |
| Hosts A, B, C with LGO | 5.4 to 5.9 M | 0.939 to 0.973 (0.992 to 0.997) | 0.835 to 0.917 (0.924 to 0.973) | 0.157 to 0.196 / 0.284 to 0.373 | 11.3 to 13.3 s |
| GeoTransolver with LGO (new) | 6.02 M | 0.952 (0.994) | 0.870 (0.945) | 0.213 / 0.345 | 10.7 s |
Layouts on which each of our hosts with the branch beats Transolver on full-field :
| Host A | Host B | Host C | |
|---|---|---|---|
| , 21 layouts | 18 | 20 | 11 |
| , 25 layouts | 22 | 23 | 24 |
The reviewer's expectation holds for test layouts close to the training layouts: on the split the attention models match our hosts with the branch, and Host C with the branch beats Transolver on only 11 of 21 layouts. It does not hold for layouts far from them. On the split the attention models' near-jet falls to between 0.51 and 0.58, while our three hosts with the branch stay between 0.92 and 0.97. Their near-jet relative rises from 0.085 to 0.596 (Transolver), from 0.112 to 0.641 (Transolver++) and from 0.149 to 0.657 (GeoTransolver), against a rise from 0.053–0.074 to 0.159–0.234 for our hosts with the branch.
Times are for one full field of 566,700 points on one GPU (our hosts from Appendix E; the other rows from one later timing job, which reproduces Transolver's 2.6 s). The branch is not free: our hosts with it take 11.3 to 13.3 s, about 4 to 5 times Transolver, and it makes GeoTransolver about 8 times slower (1.3 s to 10.7 s). Every model here is still more than 120 times faster than the 28-minute mean CFD solve.
Response to Reviewer dJhB (3/4)
R3-W2, continued. Because the branch is an add-on, these models can also be hosts. We added it to GeoTransolver in the same way as to Hosts A and C, and it is bit-identical to GeoTransolver at initialisation. On the test split, GeoTransolver's rises from 0.524 to 0.870, better on 24 of 25 layouts, and its near-jet from 0.513 to 0.945, better on all 25. The branch adds 1% to its parameters.
How these models differ from the branch. GeoTransolver comes closest, since it gathers boundary points in balls around each query. Two differences remain: in the PhysicsNeMo code the gathered points enter in absolute coordinates, which we checked by shifting the whole input, and they are not separated by boundary type. Transolver and Transolver++ have no neighbourhood around each query.
Would an explicit distance to the nearest supply vent be enough for an attention model? We have three pieces of evidence, each limited. On ShapeNet-Car, an explicit distance-to-boundary input cuts the host's error by 11.9% but does nothing once the branch is present, and the host with that input (0.0792 relative ) stays well behind the host with the branch (0.0515). On our data, giving Host A an unsigned distance to the nearest supply vent gave no gain (validation , single seed: full-field 0.684, against 0.699 to 0.753 for Host A over three seeds). And the per-type control in our reply to Reviewer qmPS shows that seeing the right points matters as well as how many. We also gave Transolver this input directly: a per-point distance to the nearest supply vent as an extra channel (+512 parameters, same training and learning rate, one test pass). It did not help. Test full-field went from 0.967 to 0.963 at and from 0.582 to 0.512 at , where it was better on only 8 of 25 layouts; our hosts with the branch still beat it on 24 of 25.
We did not include AB-UPT, because its input interface does not fit this task. It was built for external aerodynamics with a single surface, and its public implementation takes point coordinates only, with no per-point input for boundary type or boundary values. Our task is defined by those values: which boundary points are supply vents, and the velocity and temperature of the supply air. Running it here would require redesigning its input path, so the result would no longer be AB-UPT as published. We discuss it in Related Work, and we can add an adapted version if the reviewer considers it useful.
Thanks for the detailed review (1/4)
We thank the reviewer for the detailed review. Several points come from information that the paper should have stated, and the revision now states it. Line numbers refer to the revised PDF. Host A is Geom-DeepONet, Host B is GINOT (cross-attention) and Host C is GINO (a graph kernel with an FNO).
R3-W1: "Authors claim that "Neural operators efficiently model global flow fields but typically lack explicit local boundary information at each query point.". However, neural operators usually receive boundary information as input in the form of mask, representing, for instance, fluid node, wall node, source node, etc. [...] Authors do not clarify if they included this information as input to their "Host" models, and if so, in which particular way they did it."
Yes, every host receives this information, and the paper should have said so. Section 4 now has a paragraph on what every model receives (line 309), and we revised the sentence you quote from the abstract (lines 15 to 16), which now says that neural operators typically do not give each query point the geometry of nearby boundaries explicitly. Each model, with or without the branch, gets the same boundary point cloud of 15,000 points with 13 channels:
The hidden values are outputs of the CFD solve, so passing them in would give the model part of the answer. How each host reads the cloud:
So the hosts have the mask information the reviewer describes, as in MeshGraphNets. What the mask does not give a host is each query point's own view of the nearby boundary, in coordinates relative to that point. That is what the branch adds, and it is where the gains appear. Near-jet R2 on the δ≈0.97 test split (Table 3, line 379):
The hosts are not weak baselines chosen to make the branch look good. Each was tuned over the same four learning rates and trained for the same budget as every baseline, and Host C alone (0.627) is already above every attention baseline on this split (0.513 to 0.584, next comment).