{"id":"d4e8aba6-ccf6-44dd-bf14-abf25b6309c5","arxiv_id":"2606.07724","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"GTF-Net combines triplane features, AFNO spectral mixing, CNN refinement, and explicit geometric encodings to reduce relative L2 error on vehicle pressure and shear stress prediction versus prior baselines.","lead":"The paper presents GTF-Net, a neural network that builds triplane features from vehicle surface points and uses a dual AFNO-CNN backbone plus geometric query features to predict surface pressure and wall shear stress. If accurate, this offers a faster surrogate for early vehicle aerodynamic design than full CFD simulations.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Modest error reductions (0.012 / 0.011) lack variance estimates or significance tests","rationale":"The reader's weakest assumption targets architectural sufficiency. While that assumption is relevant, the immediate load-bearing risk for the central numerical claim is the absence of any reliability measure on the reported deltas themselves. This concern is prior to and independent of whether the triplane+AFNO+CNN design is theoretically adequate.","tokens_in":1804,"tokens_out":318,"duration_ms":17195,"concrete_test":"Re-train GTF-Net and the strongest baseline (GINO or TripNet) for five independent random seeds on the identical train/validation/test split and hyper-parameters; compute per-metric means and standard deviations. If the 95% CI for GTF-Net does not lie strictly below the baseline CI for both pressure and shear stress, the numerical claim is not supported at conventional significance levels.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim is a reduction in relative L2 error from 0.157 to 0.145 (pressure) and 0.237 to 0.226 (wall shear). These deltas are small. The provided abstract states that ablations show each component contributes, yet neither the abstract nor the implied full text supplies standard deviations across random seeds, cross-validation folds, or any statistical comparison (paired t-test, bootstrap CI) between GTF-Net and the strongest baseline. Without this, it is impossible to know whether the observed differences exceed training stochasticity.","agreement_with_reader":"disagree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes GTF-Net, a geometry-aware triplane field network for predicting aerodynamic pressure and wall shear stress on vehicle surfaces. Triplane features are constructed from sampled surface points via a shared MLP and bilinear rasterization, processed by a dual-stream AFNO-CNN backbone, and augmented at query time with directional coordinates, normal-projection features, and voxel curvature. The model is compared to Transolver, GINO, and TripNet, claiming relative L2 error reductions from 0.157 to 0.145 (pressure) and 0.237 to 0.226 (wall shear), with ablations indicating contributions from each architectural component.","tokens_in":1934,"tokens_out":428,"duration_ms":19138,"significance":"If the reported gains prove robust under rigorous validation, the work would represent a modest incremental advance in structured surrogate modeling for vehicle aerodynamics, demonstrating that explicit geometric encodings can be combined with triplane representations to capture both global coupling and local effects without full 3D volumetric input. The ablation results provide some credit for isolating component contributions.","major_comments":[{"comment":"Abstract: The central claim of outperformance rests on small absolute reductions (0.012 and 0.011) in relative L2 error, yet no standard deviations across random seeds, number of training runs, dataset size, train/test split details, or statistical significance tests (e.g., paired t-test or bootstrap CI) are supplied; this directly undermines the ability to conclude that the differences exceed training stochasticity.","section":"Abstract"},{"comment":"Abstract: The evaluation description provides no explicit statement of held-out test data usage, cross-validation procedure, or error-bar analysis, which are load-bearing for interpreting the reported improvements over the strongest baseline.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract states that ablations show each component contributes but does not quantify the per-component error deltas or reference specific tables/figures for those results.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for highlighting the importance of statistical rigor and transparent evaluation protocols in assessing the reported performance gains. We address each major comment below and commit to revisions that strengthen the manuscript without altering its core contributions.","responses":[{"response":"We agree that the absolute improvements are modest and that the lack of variability measures and formal statistical tests weakens the ability to attribute gains to the proposed architecture rather than training stochasticity. In the revised manuscript we will rerun training with at least five independent random seeds, report mean and standard deviation of relative L2 errors for both pressure and wall shear stress, explicitly state the full dataset size and the precise train/test split ratios used, and add paired t-tests (or bootstrap confidence intervals) comparing GTF-Net against the strongest baseline to establish statistical significance.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The central claim of outperformance rests on small absolute reductions (0.012 and 0.011) in relative L2 error, yet no standard deviations across random seeds, number of training runs, dataset size, train/test split details, or statistical significance tests (e.g., paired t-test or bootstrap CI) are supplied; this directly undermines the ability to conclude that the differences exceed training stochasticity."},{"response":"We acknowledge that the current abstract and evaluation section do not explicitly describe the test-set protocol or error-bar methodology. The revised version will add a clear statement confirming that all reported numbers are computed on a strictly held-out test set never seen during training or hyper-parameter selection, will detail the exact train/test split (including any stratification by vehicle geometry), and will present error bars derived from the multiple random-seed runs described in the response to the first comment.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The evaluation description provides no explicit statement of held-out test data usage, cross-validation procedure, or error-bar analysis, which are load-bearing for interpreting the reported improvements over the strongest baseline."}],"tokens_in":1443,"tokens_out":437,"duration_ms":16058,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing to know is that GTF-Net reports relative L2 error reductions of 0.012 on pressure and 0.011 on wall shear versus the strongest baseline, but those deltas are small and come without any reported standard deviations or statistical tests.\n\nThe architecture is a new combination for this task: triplanes built from surface points via shared MLP and bilinear rasterization, fed into a dual AFNO-CNN backbone, then queried with added geometric features (directional coordinates, normal projection, voxel curvature). Ablations indicate each element helps, and the comparisons to Transolver, GINO, and TripNet are relevant. The approach tries to handle both long-range flow coupling and local geometry without full volumetric input, which is a practical direction for vehicle design surrogates.\n\nThe soft spots are the lack of training details, dataset size, cross-validation, or error-bar analysis in the abstract. The improvements could easily be sensitive to hyperparameters or splits, and the central assumption that these query encodings plus the backbone capture the needed effects rests on the reported numbers alone. If the full paper has more on reproducibility or larger gains under different conditions, that would change the picture.\n\nThis is for people building ML surrogates for automotive CFD who want incremental architecture tweaks on surface fields. A reader already working with triplanes or AFNO might pick up useful implementation notes.\n\nIt deserves peer review so the methods section and any additional experiments can be examined directly.","headline":"Small error drops on surface pressure and shear but no variance or significance checks make the gains hard to trust.","tokens_in":2452,"tokens_out":362,"would_cite":false,"duration_ms":13815,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"GTF-Net builds triplane features from vehicle surface points and processes them with a dual-stream AFNO-CNN backbone plus explicit geometric encodings to predict pressure and wall shear stress more accurately than prior baselines.","keywords":["vehicle aerodynamics","triplane representation","pressure prediction","wall shear stress","neural operator","geometry encoding","surrogate modeling","CFD acceleration"],"falsifier":"An experiment on an unseen vehicle shape in which the relative L2 error for pressure or wall shear stress fails to improve over the strongest baseline or in which removing any one of the three components (AFNO mixing, CNN refinement, or geometric encodings) produces no measurable accuracy drop.","tokens_in":2714,"feed_emoji":"🚗","tokens_out":837,"duration_ms":15313,"temperature":0.7,"pith_summary":"The paper presents GTF-Net as a method to predict aerodynamic fields on vehicle surfaces by first constructing triplane features directly from sampled surface points using a shared MLP followed by bilinear rasterization. These planes are handled by a backbone that mixes adaptive Fourier neural operator spectral processing for long-range flow coupling with CNN refinement for local details. At query time the model adds directional coordinates, normal projections, and a voxel curvature proxy to incorporate geometry information. The result is a surrogate that avoids full 3D volumetric input yet lowers relative L2 error on pressure and shear stress compared with Transolver, GINO, and TripNet. Ablations confirm that each added component contributes measurably to the observed accuracy gain.","feed_headline":"Triplane network lowers vehicle pressure error to 0.145","feed_subtitle":"GTF-Net combines surface-point triplanes, AFNO-CNN processing, and geometric encodings to beat prior baselines on aerodynamic fields without","key_machinery":"The geometry-aware triplane field network (GTF-Net) that constructs triplane features from surface points via shared MLP and rasterization, processes them with dual-stream AFNO spectral mixing plus CNN refinement, and augments queries with directional, normal-projection, and curvature encodings.","core_discovery":"Constructing triplane features directly from sampled surface points through a shared MLP and smooth bilinear rasterization, then processing the planes with a dual-stream AFNO-CNN backbone and augmenting queries with vehicle-aligned directional coordinates, normal-projection features, and voxel-based curvature, produces surface pressure and wall shear stress predictions whose relative L2 errors are 0.145 and 0.226 respectively, lower than the strongest baseline values of 0.157 and 0.237.","pith_inferences":["The same triplane-plus-geometry encoding pattern could be tested on non-vehicle surfaces such as aircraft or wind-turbine blades to check transferability.","If the method scales, it could be inserted into shape-optimization loops that currently rely on repeated CFD runs.","The dual-stream backbone might be adapted to other surface-governed physics problems such as structural stress or heat flux prediction.","Further work could measure wall-clock speedup against full CFD on identical hardware to quantify the practical design-cycle gain."],"forward_implications":["Pressure prediction reaches a relative L2 error of 0.145 versus 0.157 for the best prior method.","Wall shear stress prediction reaches a relative L2 error of 0.226 versus 0.237 for the best prior method.","Ablation tests show that AFNO spectral mixing, local CNN refinement, and query-side geometric encoding each contribute to the accuracy improvement.","The structured triplane representation combined with explicit aerodynamic geometry cues supports accurate surface-field prediction without requiring full volumetric simulation."],"fun_headline_variants":["GTF-Net triplanes predict vehicle pressure at 0.145 L2 error","Surface-sampled triplanes achieve 0.226 wall shear L2 error","AFNO-CNN processes triplane features for vehicle aero fields","Vehicle-aligned coordinates aid triplane predictions at 0.145 error"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The combination of shared-MLP triplane construction, dual-stream AFNO-CNN backbone, and query-time geometric encodings is sufficient to capture both global aerodynamic coupling and local geometry effects without full 3D volumetric input or explicit flow physics.","fun_headline_variants_meta":{"raw":{"variants":["GTF-Net triplanes predict vehicle pressure at 0.145 L2 error","Surface-sampled triplanes achieve 0.226 wall shear L2 error","AFNO-CNN processes triplane features for vehicle aero fields","Vehicle-aligned coordinates aid triplane predictions at 0.145 error"]},"model":"grok-4.3","cost_usd":0.00613,"raw_usage":{"total_tokens":2852,"prompt_tokens":746,"num_sources_used":0,"completion_tokens":77,"cost_in_usd_ticks":61303000,"prompt_tokens_details":{"text_tokens":746,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2029,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":746,"tokens_out":77,"duration_ms":14621,"temperature":1.0,"reasoning_tokens":2029,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-27T22:40:59.300486+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"An experiment on an unseen vehicle shape in which the relative L2 error for pressure or wall shear stress fails to improve over the strongest baseline or in which removing any one of the three components (AFNO mixing, CNN refinement, or geometric encodings) produces no measurable accuracy drop.","supporting_citations":[],"review_version":1}