{"id":"898b20c7-db2f-4f9b-9d34-79aa05b84b8c","arxiv_id":"2607.20846","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A fixed D3Q125 kinetic model with separate per-moment-order log-Gaussian relaxation reduces nonequilibrium relative to a one-sensor model, but the gain is configuration-dependent and not externally validated.","lead":"This paper presents a way for a fixed D3Q125 discrete-velocity kinetic solver to relax second-, third-, and fourth-order moment deviations with three separate sensors instead of one shared scalar signal. In one benchmark the approach cuts peak nonequilibrium by 6.565%, but only against the author's own baseline and without independent kinetic-reference validation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The λ=0 peak-TNE reduction is a monotonicity artifact of comparing an additive common sensor (Etot) with component-wise sensors (En); the sign is mathematically guaranteed, so the headline number does not by itself support the order-resolved mechanism.","rationale":"The reader's CONDITIONAL verdict is well supported. My stress-test sharpens the reader's concern about 'designed consequence' into a precise internal point. In the λ=0 limit, Kcommon reduces to Etot=E2+E3+E4, while K_n=En. Since s_n(K) is strictly decreasing in K (erfc and s_cont>s_kin), each resolved sector has a larger relaxation factor than the common sector, so all three sector TNE values are reduced. This is not an empirical observation; it follows from the definitions in §3.4–3.5. The sign robustness across spectrum perturbations is therefore also expected, provided monotonicity and s_cont>s_kin hold. The paper is transparent about not claiming accuracy and about parameter sensitivity, and the mechanism/selectivity tests are internally consistent. However, the headline 6.565% is best read as a property of the chosen comparator, not as evidence that order-resolved activation is physically better. The concrete test with a non-additive common sensor would show whether any advantage survives a fairer scalar comparator; an independent kinetic reference would be needed for accuracy. Since the reader already conditioned acceptance on exactly this class of missing validation, no further verdict change is needed.","tokens_in":25056,"tokens_out":6878,"duration_ms":75658,"concrete_test":"Recompute the λ=0 smooth compression-wave comparison replacing the common-sensor definition with a non-additive scalar sensor, e.g. Kcommon = N_8(E2,E3,E4) = (E2^8+E3^8+E4^8)^{1/8} or Kcommon = max(E2,E3,E4), and report the peak-total-TNE ratio. If the order-resolved model no longer reduces peak TNE, the reported 6.565% is an artifact of the additive common-sensor definition; if the reduction persists under a non-additive scalar sensor, the order-resolved mechanism has content beyond monotone splitting.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing problem is not the uncalibrated spectrum itself but that the headline λ=0 comparison is an analytical consequence of the comparator choice. From §3.4–3.5, in the TNE-only limit Kρ=KT=Ku=0, so the common sensor is Kcommon=N_p(0,0,0,Etot)=Etot=E2+E3+E4, whereas the order-resolved sensors are Kn=En for n=2,3,4. The relaxation factor s_n(K)=wc,n(K)·1+(1−wc,n(K))·sn,kin is strictly decreasing in K because erfc decreases and 1>sn,kin. Hence, whenever at least two sectors are nonzero, s_n^resolved>s_n^common for every n, so every sector is relaxed more strongly and every sector's TNE is reduced. The sign of the reduction is therefore guaranteed by monotonicity for any spectrum with s_cont>s_kin; the only freedom is the magnitude, set by the hand-chosen K0,n, σn, sn,kin. Thus the 6.565% result demonstrates that splitting an additive scalar into components before a monotone map lowers the output; it does not by itself demonstrate that order-resolved activation is physically preferable. The paper honestly disclaims accuracy, but the central quantitative claim remains essentially built into the construction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an order-resolved hierarchical collision model on a fixed D3Q125 discrete-velocity set. A shared macroscopic-gradient rarefaction background is combined separately with second-, third-, and fourth-order thermodynamic nonequilibrium (TNE) measures to form three effective Knudsen indicators K2, K3, K4, each driving its own log-Gaussian relaxation spectrum. Numerical tests cover pure-order activation, homogeneous mixed-order relaxation, amplitude and composition scans, a temperature wave, a smooth compression wave, grid/timestep sensitivity, transport-discretization sensitivity, relaxation-spectrum sensitivity, uniform-boost frame checks, and a shear-wave modal-decay study. The headline result is that in the TNE-only sensor limit (λ=0) of the compression-wave benchmark, the order-resolved model reduces peak total TNE by 6.565% relative to the common-sensor model, with reductions in every retained moment sector. The paper explicitly disclaims universal accuracy improvement and identifies external kinetic-reference validation as the principal remaining step.","tokens_in":25533,"tokens_out":6756,"duration_ms":69259,"significance":"The paper is careful, well structured, and unusually honest about its limitations. The implementation checks are extensive and convincing as numerical verifications of the stated update rules: conservation to floating-point accuracy, positivity, long-time runs, pure-order selectivity, and quantitative boost diagnostics are all reported. However, the central numerical comparison is largely forced by the construction: with the chosen monotone relaxation schedule, splitting a common scalar sensor into component-wise sensors guarantees larger relaxation factors and hence lower residual TNE. Thus the 6.565% number and the sign of the reduction are not independent empirical evidence for a physical advantage of order resolution. The useful contribution is the explicit, well-tested hierarchical construction and the honest mapping of its regime dependence, not the quantitative accuracy claim. The paper would be acceptable as a mechanism/numerical-behavior study after the forced nature of the headline result is stated and the interpretation is reframed accordingly.","major_comments":[{"comment":"The λ=0 reduction is mathematically guaranteed by the definitions. With Kρ=KT=Ku=0, the sensors reduce to Kn=En and Kcommon=Etot. Since Etot>En whenever at least two TNE sectors are nonzero, and since s_n(K) is strictly decreasing in K (the erfc weight decreases with K and s_cont>s_kin), we have s_n^resolved>s_n^common for every n. Each sector is therefore relaxed more strongly, and the sector-wise and total TNE reductions follow automatically. The same monotonicity argument underlies the homogeneous results in §5.2–5.3 and the sign-robustness claims in §5.11–5.13. The paper should state this analytical guarantee explicitly and reframe the 6.565% result as a consequence of the comparator choice, not as an empirical discovery about the value of order resolution.","section":"§3.4 and §5.5.3"},{"comment":"The quantitative magnitude of the reported reductions is controlled by uncalibrated parameters (K0,n, σn, sn,kin) taken from the author's prior continuum-ballistic preprint [20], together with the selected norm order p=8 and sensor coefficients. The paper's own sensitivity scans show the reduction varying from 16.47% to 21.50% (spectrum scan, CFL=0.05) and from 2.418% to 11.523% (grid/timestep scan), so the 6.565% value is not a physical constant. Since no independent kinetic reference is provided, the numerical results cannot support any accuracy claim. The abstract and conclusions should consistently present these numbers as illustrative of the construction and should not imply that the model is validated by the internal common-sensor comparison.","section":"§3.5, Table 3, §5.13"},{"comment":"The laboratory-frame raw-Hermite sensor is not Galilean invariant. The uniform-boost tests show that a boost alone changes the third- and fourth-order TNE and the relaxation factors (for example, s4 drops from 1 to 0.489 at U0=0.4 in the homogeneous test, and transport-enabled peak TNE increases by 50.6% at U0=0.2). The paper acknowledges this and provides quantitative diagnostics, which is commendable. Nevertheless, for a kinetic model intended for compressible or high-speed flow this frame dependence is a genuine physical deficiency. At minimum, the claims should be restricted to the stated frame/mean-flow regime, and a central-moment or central-Hermite formulation should be identified as necessary for a frame-independent sensor.","section":"§5.10 and §6.8"}],"minor_comments":[{"comment":"The Common-model column in the grid-and-timestep table appears to have row-count digits accidentally prepended: the listed values like 11.226703×10−2, 26.394078×10−3, 43.653046×10−3, and 82.384581×10−3 contradict the text and the computed percentages. They should presumably read 1.226703×10−2, 6.394078×10−3, 3.653046×10−3, and 2.384581×10−3.","section":"§5.5.5, Table"},{"comment":"Text such as 'CFL= 0.4to0 .0015625' contains spacing and formatting errors; please fix for clarity.","section":"§5.9, Table 6"},{"comment":"The notation Sym for normalized full index symmetrization is introduced without an explicit definition of the normalization factor. Please define it precisely.","section":"§5.10"},{"comment":"The paper would benefit from an explicit statement about code and data availability. The text states that quadrature abscissae are generated by the implementation, but no repository or reproducibility artifact is mentioned.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is competently written and unusually transparent about limitations, but the central numerical claim is analytically forced by the monotone sensor construction. I recommend major revision rather than rejection because the model and numerical implementation are sound and the authors already disclaim universal accuracy. The revision must explicitly derive the monotonicity argument, reframe the 6.565% result as a property of the comparator, and clearly state that the quantitative values are uncalibrated and not evidence of physical accuracy. Without that reframing, the headline result is circular."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my take. If you only have two minutes: the model construction is real and the numerical work is careful and honest, but the headline number (6.565% peak-TNE reduction) does not give the order-resolved mechanism any independent support, because its sign is mathematically guaranteed by the choice of comparator.\n\nWhat is new: the author combines a shared macroscopic-gradient indicator with separate second-, third-, and fourth-order raw-Hermite TNE measures, each driving its own log-Gaussian relaxation curve on a fixed D3Q125 set. Pure-order activation, conservation checks, positivity, long-time runs, timestep, transport, spectrum, and frame-dependence sensitivity studies are all well executed and clearly reported. The shear-wave modal-decay and mesh-refinement section is a useful extra. The paper also deserves credit for explicitly flagging that it is not claiming universal accuracy, that the spectrum is a research baseline, and that the lab-frame sensor has boost sensitivity.\n\nThe soft spot is the one the stress-test note names, and it is load-bearing. In the λ=0 limit the common sensor is Etot=E2+E3+E4 and the order-resolved sensors are En. The relaxation factor s_n(K) decreases with K. Therefore each sector gets a larger s (stronger relaxation) in the resolved model than in the common model whenever at least two channels are nonzero. The sign of the TNE reduction is guaranteed for any monotone schedule with s_cont>s_kin; only the magnitude depends on the hand-set parameters. So the 6.565% number demonstrates that splitting an additive scalar before a monotone map lowers the output; it does not show that order-resolved activation is physically preferable. The author says the right words about this in Section 6, but the central quantitative claim is still essentially baked into the construction.\n\nAdditionally, there is no independent kinetic reference, no code or data shipped, and all comparisons are against the author's own baseline. These are acknowledged, but they mean the paper's value is as a mechanism demonstration, not a validated improvement.\n\nWho it is for: people working on high-order LBM/DBM collision models who want an example of per-order sensor construction. It deserves a serious referee because the numerics are thorough and the framing is honest; but the referee should push for a reframed conclusion and external validation before acceptance. I would not cite it in my own work yet.","headline":"Clear, honest numerical study of a new order-resolved collision construction, but the headline 6.565% reduction is a built-in monotonicity consequence of comparing En vs Etot sensors, not independent evidence of physical advantage.","tokens_in":26008,"tokens_out":4032,"would_cite":false,"duration_ms":37981,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76P05","82C40","65M08"],"pacs":[],"model":"deepseek-v4-flash","headline":"Decoupling moment-order relaxation cuts peak thermodynamic nonequilibrium by 6.565%","keywords":["hierarchical collision model","order-resolved relaxation","thermodynamic nonequilibrium","D3Q125 velocity set","log-Gaussian relaxation spectrum","effective Knudsen indicator","discrete-velocity kinetic model","adaptive rarefaction sensor"],"falsifier":"Run the λ=0 smooth compression-wave benchmark with an independent high-resolution discrete-velocity or DSMC solver as reference: if the order-resolved peak total TNE moves closer to the reference than the common-sensor peak does, the central claim is supported; if it moves further away, the reduction is a numerical artifact. A second falsifier is a spectrum-parameter scan outside the tested range that reverses the sign of the total-TNE difference, which would break the claimed robustness of the correction's sign.","tokens_in":24884,"feed_emoji":"⚛️","tokens_out":10423,"duration_ms":97263,"temperature":0.7,"pith_summary":"Adaptive collision models in discrete-velocity kinetic methods usually compute one scalar rarefaction or nonequilibrium indicator and feed it to every retained moment order, coupling sectors that may be far from equilibrium at different times and places. This paper develops and tests a hierarchical alternative: a shared macroscopic-gradient background combined separately with second-, third-, and fourth-order thermodynamic nonequilibrium measures yields three effective indicators K2, K3, K4, each driving its own log-Gaussian relaxation spectrum on a fixed D3Q125 velocity set. The authors show that the resulting activation is selective (pure-order perturbations leave nonmatching sectors at roundoff), consistently lowers residual nonequilibrium relative to the common-sensor model in homogeneous and wave tests, and cuts peak total nonequilibrium by 6.565% in a smooth compression wave when only the nonequilibrium channels drive the sensor. The paper does not claim universal accuracy: the improvement is measured against its own comparator, the spectrum parameters are prescribed baselines, and independent kinetic-reference validation is declared necessary before a universal accuracy claim.","feed_headline":"Decoupling moment-order relaxation cuts peak nonequilibrium by 6.565%","feed_subtitle":"Separate sensors for stress, heat-flux, and higher-order moments beat one shared sensor in TNE-dominated regimes.","key_machinery":"The machinery is an order-resolved effective Knudsen indicator K_n = N_8(K_ρ, K_T, K_u, E_n), an unnormalized p-norm with p=8 that merges a shared macroscopic-gradient background (normalized density, temperature, and velocity gradients with scale λ) with an order-specific thermodynamic nonequilibrium measure E_n formed from the Frobenius norm of the nth-order raw-Hermite coefficient deviation from equilibrium, scaled by pressure and temperature. Each K_n enters a log-Gaussian scale-space spectrum w_{c,n}(K) = ½ erfc( ln( max(K,10⁻¹⁴)/K_{0,n} ) / (√2 σ_n) ), which blends a continuum-limit relaxation factor (s=1) with a kinetic-limit factor (s_kin = 0.20/0.10/0.05 for orders 2/3/4) and is mapp","core_discovery":"The paper's central claim is that order-resolved activation works as designed: replacing one shared nonequilibrium sensor with three order-specific effective Knudsen indicators removes the artificial coupling in which one strongly nonequilibrium sector suppresses relaxation in another. In the TNE-only sensor limit (λ=0) of a smooth periodic compression wave, peak total nonequilibrium intensity falls from 3.653046e-3 to 3.413224e-3 (6.565%), at every cell and in every retained moment sector, with the largest local relaxation correction in the fourth-order channel (Δs4 ~0.237). The correction's sign persists across timestep, transport-scheme, spectrum, boost, long-time, and shear-wave tests; i","pith_inferences":["If an independent kinetic reference confirms the direction of the correction, the same sensor-to-relaxation decoupling should transfer to other fixed velocity sets and to regularized or multiple-relaxation-time collision models, since the change is localized in the indicator construction, not in the velocity representation.","The fourth-order sector being the most sensitive suggests that retaining moments beyond fourth order would amplify the scalar-aggregation problem; the benefit of order-resolved activation may grow as more Hermite orders are kept.","A testable extension is coupling the hierarchical sensor with dynamic velocity-space adaptation (switching among D3Q125, D3Q343, and D3Q729); if the two improvements are independent their benefits should roughly add, and if they interact the coupling would reveal which mechanism actually governs accuracy.","Calibrating the log-Gaussian spectrum parameters (K0,n, σ_n, s_kin) against transport coefficients or reference solutions is the direct route to converting the qualitative robustness into a quantitative statement; absent that, the reported percentages should be read as properties of the chosen baseline schedule."],"forward_implications":["In TNE-dominated flow regimes, collision models that use one scalar sensor will over-suppress relaxation of weakly nonequilibrium sectors; order-resolved activation removes that coupling, so similar smooth-flow tests should show systematically lower residual TNE.","The reported 6.565% and related percentages are configuration-dependent: timestep, transport scheme, and the prescribed spectrum all change the magnitude, so quantitative comparisons between studies must control these settings.","Because the method preserves conserved moments to floating-point accuracy, keeps populations positive without limiting in the main benchmarks, and adds only about 2% runtime in the reference implementation, it can be adopted into existing finite-volume discrete-velocity codes without changing the velocity set or transport stage.","As the macroscopic-gradient background strengthens, the common and order-resolved models converge (to order 10⁻⁷ in local factor differences at λ=0.04), meaning the hierarchical correction matters mainly in regions where nonequilibrium channels carry information the scalar sensor would discard.","The documented frame dependence of the raw-Hermite sensor under uniform boosts identifies central or locally centered Hermite moments as the natural next step; the paper shows the effect follows the analytical translation structure and is not an implementation artifact over the tested range."],"fun_headline_variants":["Order-specific sensors cut peak nonequilibrium by 6.565%","Decoupling moment orders trims TNE intensity 6.565%","Three Knudsen sensors beat one shared sensor","Separate moment sensors reduce peak TNE by 6.565%","Order-resolved relaxation lowers peak TNE 6.565%"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The comparison that supports the claimed advantage uses the paper's own common-sensor model as the only comparator and a hand-set log-Gaussian spectrum as the only relaxation schedule; if those prescribed parameters are arbitrary, the 6.565% reduction is likewise arbitrary, and without an independent kinetic reference the reduction cannot be tied to physical accuracy.","fun_headline_variants_meta":{"raw":{"variants":["Order-specific sensors cut peak nonequilibrium by 6.565%","Decoupling moment orders trims TNE intensity 6.565%","Three Knudsen sensors beat one shared sensor","Separate moment sensors reduce peak TNE by 6.565%","Order-resolved relaxation lowers peak TNE 6.565%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000608,"raw_usage":{"total_tokens":2721,"prompt_tokens":851,"completion_tokens":1870,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":1794}},"tokens_in":595,"tokens_out":1870,"duration_ms":13161,"temperature":1.0,"reasoning_tokens":1794,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T09:08:22.643610+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the λ=0 smooth compression-wave benchmark with an independent high-resolution discrete-velocity or DSMC solver as reference: if the order-resolved peak total TNE moves closer to the reference than the common-sensor peak does, the central claim is supported; if it moves further away, the reduction is a numerical artifact. A second falsifier is a spectrum-parameter scan outside the tested range that reverses the sign of the total-TNE difference, which would break the claimed robustness of the correction's sign.","supporting_citations":[],"review_version":1}