{"id":"93f27b93-3cb8-49a3-86f3-87a6c684744c","arxiv_id":"2606.00949","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A combined SHAP-guided MARL strategy using U-net predictions of skin-friction and wall pressure achieves 34.44% drag reduction and 34.01% net energy saving with 0.43% normalized input power in turbulent channel flow.","lead":"The paper trains multi-agent reinforcement learning agents guided by SHAP explanations from U-nets to control near-wall turbulence and reduce drag. A smart generalist might read it to see how explainable AI can discover efficient physical control rules that outperform hand-designed strategies in fluid systems.","discovery_kind":"unclear","skeptic_critique":{"model":"grok-4.3","headline":"U-net prediction accuracy and whether SHAP attributions yield a causally effective reward rather than spurious correlations","rationale":"The reader's weakest_assumption correctly isolates the single point where the argument is least secured. Because the full text was not supplied in the query, no additional internal inconsistency or stronger concern could be located; the surrogate-model risk remains the dominant one.","tokens_in":1843,"tokens_out":404,"duration_ms":16190,"concrete_test":"Hold out 20% of the flow snapshots; retrain or evaluate the two U-nets and report MSE / Pearson correlation for 1-, 5-, and 10-step-ahead predictions of skin-friction and wall pressure. If either metric falls below 0.85 correlation or MSE exceeds 15% of the signal variance, recompute the SHAP rewards on the held-out set and re-train the MARL agents; if the 34% DR figure drops by more than 5 percentage points, the performance claim is sensitive to surrogate-model error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline performance numbers (34.44% DR, 34.01% NES at 0.43% power) rest on the three SHAP-guided reward formulations. These rewards are generated by applying SHAP to U-nets that predict future skin-friction coefficient and/or wall-pressure fluctuations from the instantaneous flow state. For the reported gains over opposition control and the direct wall-shear baseline to be attributable to the method, two conditions must hold: (1) the U-nets must generalize accurately enough that their attributions identify control-relevant features, and (2) optimizing the policy against those attributions must produce long-term improvement in the true Navier-Stokes dynamics rather than exploitation of model error. The manuscript provides no quantitative validation (test-set MSE, correlation, or horizon-dependent error) of the U-nets nor an ablation that replaces the SHAP reward with the ground-truth future quantities.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes a hybrid MARL+XDL framework for turbulent drag reduction in wall-bounded flows. Three SHAP-guided reward formulations are derived from U-nets that predict future velocity, skin-friction coefficient, or wall-pressure fluctuations; the best-performing variant (combined skin-friction and pressure SHAP) is reported to achieve 34.44% drag reduction and 34.01% net energy saving at 0.43% normalized input power, outperforming both opposition control and a direct wall-shear-stress baseline while lowering actuation cost.","tokens_in":2040,"tokens_out":496,"duration_ms":11989,"significance":"If the U-net predictions generalize and the resulting SHAP rewards produce policies that improve the true Navier-Stokes dynamics, the work would supply both a high-performance, low-power control law and an interpretable link between near-wall pressure and actuation timing. The explicit comparison against two external baselines and the reported energy metrics would constitute a concrete advance in data-driven flow control.","major_comments":[{"comment":"The central performance claims (34.44% DR, 34.01% NES) rest on SHAP attributions obtained from U-nets whose predictive accuracy is never quantified. No test-set MSE, correlation coefficient, or horizon-dependent error is reported for any of the three U-nets, making it impossible to determine whether the attributions identify causally relevant features or merely exploit model error.","section":"Abstract / Methods (U-net training)"},{"comment":"No ablation is presented that replaces the SHAP-derived reward with the ground-truth future skin-friction or pressure values. Without this control, it remains unclear whether the reported gains over the direct wall-shear baseline arise from the explainability step or simply from using a different reward formulation.","section":"Results (SHAP-guided strategies)"}],"minor_comments":[{"comment":"The abstract states relative improvements of 49.41% and 48.52% over opposition control; these percentages should be accompanied by absolute values and statistical uncertainty to allow direct comparison.","section":"Abstract"},{"comment":"Notation for normalized input power and net energy saving is introduced without an explicit equation; a short definitions subsection would improve reproducibility.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"Thank you for the opportunity to respond to the referee's comments on our manuscript. We appreciate the constructive feedback and address each major comment below. We propose revisions to strengthen the presentation of the U-net models and clarify the role of the SHAP step.","responses":[{"response":"We agree that the predictive accuracy of the U-nets must be quantified to support the reliability of the SHAP attributions. In the revised manuscript we will add a dedicated subsection (or appendix) reporting test-set MSE, Pearson correlation coefficients, and horizon-dependent error curves for all three U-nets. These metrics will confirm that the models achieve sufficient accuracy for the attributions to reflect physically relevant features rather than model artifacts.","revision_made":"yes","referee_comment":"[Abstract / Methods (U-net training)] The central performance claims (34.44% DR, 34.01% NES) rest on SHAP attributions obtained from U-nets whose predictive accuracy is never quantified. No test-set MSE, correlation coefficient, or horizon-dependent error is reported for any of the three U-nets, making it impossible to determine whether the attributions identify causally relevant features or merely exploit model error."},{"response":"The direct wall-shear-stress baseline already employs instantaneous measurements as the reward. Using ground-truth future skin-friction or pressure as a reward would require an oracle unavailable in any practical online control setting and would therefore not constitute a fair ablation of the SHAP component. The SHAP formulation is deliberately chosen to extract interpretable, predictive features from the U-net outputs. In the revision we will expand the discussion to explicitly contrast the instantaneous baseline with the predictive SHAP approach and will add a limited comparison that replaces SHAP with the raw U-net predictions (without attribution) to isolate the contribution of the explainability step.","revision_made":"partial","referee_comment":"[Results (SHAP-guided strategies)] No ablation is presented that replaces the SHAP-derived reward with the ground-truth future skin-friction or pressure values. Without this control, it remains unclear whether the reported gains over the direct wall-shear baseline arise from the explainability step or simply from using a different reward formulation."}],"tokens_in":1450,"tokens_out":473,"duration_ms":21388,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The one thing to know is that a reward built from SHAP attributions of two U-nets—one for skin-friction coefficient and one for wall-pressure fluctuations—lets the multi-agent RL agents reach 34.44% drag reduction and 34.01% net energy saving at only 0.43% normalized input power. That beats opposition control by roughly 49% on both metrics and improves on the direct wall-shear baseline while dropping actuation cost from 5.90% to 0.43%.\n\nWhat is new is the reward construction: the agents optimize against feature attributions extracted from the U-nets rather than against raw physical targets. The paper tests three SHAP variants, identifies the combined skin-friction plus pressure version as best, and adds an analysis showing the learned policy activates mainly at near-zero wall pressure on timescales that match near-wall turbulent structures.\n\nThe comparison to two external baselines is straightforward and avoids circularity. The performance numbers are stated clearly and the physical interpretation is consistent with existing turbulence literature.\n\nThe soft spot is the complete absence of U-net validation. The abstract gives no test-set error, no prediction-horizon accuracy, and no ablation that replaces the SHAP reward with ground-truth future quantities. Without those checks it remains possible that the reported gains come from the agents exploiting surrogate-model errors rather than from genuinely better control features. The stress-test note flags exactly this gap, and the abstract does not close it.\n\nThis work is for researchers at the overlap of reinforcement learning and wall-bounded turbulence control. Readers already running MARL on channel flow will see a concrete alternative reward design worth testing. It deserves peer review because the numerical claims are specific, the method is distinct from the cited baselines, and a referee can request the missing U-net diagnostics and any error bars or statistical tests.","headline":"SHAP rewards from dual U-nets in MARL cut actuation cost while beating opposition control and direct shear baselines on drag reduction, but the U-nets lack any reported accuracy checks.","tokens_in":2550,"tokens_out":454,"would_cite":false,"duration_ms":20862,"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":"SHAP attributions from U-net predictors shape MARL rewards to discover a pressure-gated policy that reduces turbulent drag by 34% while using only 0.43% input power.","keywords":["turbulent drag reduction","deep reinforcement learning","SHAP explainability","multi-agent RL","wall-bounded turbulence","skin-friction coefficient","wall-pressure fluctuations","energy-efficient control"],"falsifier":"Deploying the trained agents in an independent direct numerical simulation at the same Reynolds number and observing whether the measured drag reduction falls below 30% or the normalized actuation power rises above 1% would falsify the performance claims.","tokens_in":2735,"feed_emoji":"🌊","tokens_out":782,"duration_ms":21106,"temperature":0.7,"pith_summary":"The paper establishes that incorporating SHAP values from two U-nets—one predicting skin-friction coefficient and one predicting wall-pressure fluctuations—into the reward function of multi-agent deep reinforcement learning produces control policies superior to both direct wall-shear targeting and opposition control. A sympathetic reader would care because the resulting strategy simultaneously raises drag reduction and net energy saving while cutting normalized actuation cost by more than an order of magnitude. Analysis of the learned policy shows it activates mainly at near-zero wall pressure and on timescales comparable to the lifetime of near-wall turbulent structures.","feed_headline":"SHAP rewards cut turbulent drag by 34% at 0.43% power cost","feed_subtitle":"Combined skin-friction and pressure attributions produce a pressure-gated policy that beats opposition control while slashing actuation cost","key_machinery":"SHAP attributions from U-nets predicting future skin-friction coefficient and wall-pressure fluctuations, used as the reward signal for multi-agent reinforcement learning agents.","core_discovery":"The combined SHAP strategy based on skin-friction coefficient and wall-pressure fluctuations achieves the best overall performance, achieving a DR of 34.44% and a NES of 34.01% with only 0.43% normalized input power. Relative to opposition control, drag reduction and net energy saving increase by 49.41% and 48.52%, respectively. Compared with the direct wall-shear-stress baseline, the proposed strategy simultaneously improves performance while reducing the normalized actuation cost from 5.90% to 0.43%. The energetically efficient policy is consistent with pressure-gated actuation, activating predominantly at near-zero wall pressure, and operates on a temporal timescale comparable to the life","pith_inferences":["The same SHAP-reward construction could be tested on other canonical wall-bounded flows such as pipe or boundary-layer turbulence to check transferability.","Laboratory experiments that implement only pressure-based actuation at near-zero crossings would provide a direct physical test of the simulated energy savings.","Future sensor designs for active flow control might prioritize wall-pressure measurements over shear-stress measurements if the pressure-gated mechanism holds."],"forward_implications":["The pressure-gated policy achieves 34.44% drag reduction and 34.01% net energy saving at 0.43% normalized input power.","Actuation occurs predominantly when instantaneous wall pressure is near zero.","The temporal scale of effective control matches the lifetime of near-wall turbulent structures.","Normalized actuation cost drops from 5.90% in the direct wall-shear baseline to 0.43% while performance improves."],"fun_headline_variants":["SHAP-guided MARL achieves 34% drag reduction at 0.43% power cost","Combined skin-friction pressure SHAP achieves 34% drag reduction","Pressure-gated actuation from SHAP achieves 34% drag reduction","XDL MARL strategy achieves 34% drag reduction at 0.43% cost"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The U-nets accurately predict future skin-friction and pressure values from the current flow state so that their SHAP attributions produce a reward signal that genuinely improves the agents' long-term control policy rather than rewarding spurious correlations.","fun_headline_variants_meta":{"raw":{"variants":["SHAP-guided MARL achieves 34% drag reduction at 0.43% power cost","Combined skin-friction pressure SHAP achieves 34% drag reduction","Pressure-gated actuation from SHAP achieves 34% drag reduction","XDL MARL strategy achieves 34% drag reduction at 0.43% cost"]},"model":"grok-4.3","cost_usd":0.007779,"raw_usage":{"total_tokens":3609,"prompt_tokens":780,"num_sources_used":0,"completion_tokens":83,"cost_in_usd_ticks":77787000,"prompt_tokens_details":{"text_tokens":780,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2746,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":780,"tokens_out":83,"duration_ms":18774,"temperature":1.0,"reasoning_tokens":2746,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T17:57:29.488651+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Deploying the trained agents in an independent direct numerical simulation at the same Reynolds number and observing whether the measured drag reduction falls below 30% or the normalized actuation power rises above 1% would falsify the performance claims.","supporting_citations":[],"review_version":1}