{"id":"2a84d310-990b-427f-b131-134194739166","arxiv_id":"2510.26276","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":6,"one_line_summary":"A Bayesian centrality method using a two-dimensional Gamma kernel on spectator energy and track hits claims to reproduce BM@N simulated impact-parameter distributions within 2%, but contains a load-bearing rotation error and self-consistent validation.","lead":"The paper proposes a two-dimensional Bayesian method for determining how central a heavy-ion collision is, combining spectator energy and track-hit number through a product of two Gamma distributions. It reports that the method reproduces simulated BM@N impact-parameter distributions within 2%, but the central formula has an error and the validation is in-sample.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rotation angle in §IV is missing the 1/2 factor, so Eq. (6)'s product-of-Gamma kernel is not derived and the centrality reconstruction rests on an invalid covariance diagonalization.","rationale":"The reader's weakest assumption and my independent reading converge on the same point: the rotation angle in Section IV is missing the factor 1/2 required to diagonalize the covariance matrix, so the product-of-Gamma form in Eq. (6) is not a consequence of the stated transformation. This is not a stylistic or consensus disagreement; it is an internal inconsistency in the derivation. The paper even writes sin(2phi) in Eqs. (4)-(5), which reinforces that the argument requires tan(2phi), not tan(phi). With the text's phi, a direct calculation gives Cov(X1,X2) = -Cov(Espec, Nhits), not zero. The rest of the centrality extraction—Eq. (8), the fits in Figs. 6-7, and the reconstructed impact-parameter distributions in Figs. 9-10—all build on this kernel, so the central claim is currently unsupported as written. I also note, as a secondary concern, that the validation is self-referential: mean values, variances, and covariance are extracted from the same DCM-QGSM-SMM simulation that is later used as the reference for the 2% agreement. That circularity would matter even after the angle typo is fixed. The method is likely repairable, and the paper identifies a real need for two-dimensional centrality determination at low energies, but the present text has a concrete, checkable mathematical error. Therefore I agree with the reader's REJECT verdict and recommend no change.","tokens_in":8359,"tokens_out":7407,"duration_ms":80646,"concrete_test":"Using one representative impact-parameter slice from Fig. 4 (e.g., a central and a peripheral value), plug the reported moments and covariance into Eqs. (3)-(5) with the paper's stated phi = arctan(2Cov/(sigma_E^2 - sigma_N^2)) and compute Cov(X1,X2). It will be nonzero (in fact equal to -Cov(Espec, Nhits)). Then repeat with phi = (1/2) arctan(2Cov/(sigma_E^2 - sigma_N^2)) and verify Cov(X1,X2)=0. If the text's angle is used in the actual reconstruction, the factorized kernel of Eq. (6) is not the kernel being applied; this single numerical check settles whether the derivation is correct as written.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central kernel, Eq. (6), is a product of two independent Gamma distributions in rotated variables. This factorization is justified only if Cov(X1,X2)=0 after the rotation. The paper claims that the angle phi = arctan[2Cov/(sigma_E^2 - sigma_N^2)] in Section IV achieves this. It does not. Diagonalizing a 2x2 covariance matrix requires tan(2phi) = 2Cov/(sigma_E^2 - sigma_N^2), i.e. phi = (1/2) arctan(2Cov/(sigma_E^2 - sigma_N^2)). With the paper's phi, one obtains Cov(X1,X2) = -Cov(Espec, Nhits), not zero. For example, if sigma_N^2=1, sigma_E^2=6, Cov=1, then phi=arctan(0.4)=0.3805 rad and Cov(X1,X2) ≈ -1. Therefore Eq. (6) does not follow from the stated construction. Even with the correct half-angle, the independence of the rotated variables is an unverified modeling assumption rather than a proven property of the physics. Since Eq. (6) feeds directly into Eq. (8) and all centrality-class and impact-parameter reconstructions in Section V, the quantitative claim of <2% agreement with DCM-QGSM-SMM in Fig. 10 is supported only if the implementation silently uses a different angle than the text states. The method may be repairable, but as written the load-bearing mathematical step is incorrect.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a two-dimensional Bayesian centrality-determination method for BM@N Xe+CsI collisions at 3.8A GeV. The authors propose that at fixed impact parameter the joint distribution of track-hit multiplicity and FHCal spectator energy factorizes into a product of two Gamma distributions after a rotation of the observables; the means and variances of the Gamma parameters are obtained from DCM-QGSM-SMM simulation and parameterized by polynomials in centrality. Additional scaling parameters α_E, α_N, β_E, β_N, and an efficiency ε are introduced to connect experiment and simulation. The full two-dimensional distribution is fit to the DCM-QGSM-SMM data, centrality classes are defined via constrained k-means, and the resulting impact-parameter distributions and mean-impact-parameter dependence are compared with the same model, claiming agreement within 2%.","tokens_in":8812,"tokens_out":6740,"duration_ms":67245,"significance":"The paper addresses a relevant problem for low-energy heavy-ion fixed-target experiments, where a non-monotonic FHCal response requires more than one observable for centrality determination. Its strengths include a realistic GEANT4 setup, a complete data-analysis chain, and quantitative comparisons. However, the central mathematical step contains an incorrect rotation angle, the factorized-Gamma kernel is an unverified modeling assumption, and the validation is performed in-sample on the same simulation that provides the parameterization. As a result, the paper's main quantitative claim is not currently supported. The methodological core appears repairable, but the present draft requires substantial correction and additional validation.","major_comments":[{"comment":"The rotation angle used to diagonalize the covariance matrix is incorrect. Diagonalization requires tan(2φ)=2Cov/(σ_E^2−σ_N^2), i.e. φ=(1/2)arctan(2Cov/(σ_E^2−σ_N^2)). The paper defines φ=arctan(2Cov/(σ_E^2−σ_N^2)). With this angle, Cov(X1,X2)=0.5·sin(2φ)(σ_N^2−σ_E^2)+cos(2φ)Cov, which is not generally zero. For example, with σ_N^2=1, σ_E^2=6, and Cov=1, φ≈0.380 rad and Cov(X1,X2)≈−1. Therefore Eq. (6) does not follow from the stated construction, and this kernel feeds directly into Eq. (8) and all reconstructions in Section V. The authors must correct the angle or explicitly state the transformation actually used in the numerical implementation.","section":"§IV, Eqs. (3)–(6)"},{"comment":"Even with the corrected half-angle, zero covariance does not imply independence of the rotated variables, nor does it imply that each rotated variable is Gamma distributed. The product-of-Gamma form in Eq. (6) is a modeling assumption, not a consequence of the covariance diagonalization. Since the entire Bayesian inversion uses this kernel, the assumption must be tested explicitly—for example by comparing the fitted two-dimensional density to the empirical distribution in narrow impact-parameter bins or by a quantitative goodness-of-fit test. The lower panels of Fig. 3 show contour comparisons but no statistical quantification of agreement.","section":"§IV, Eq. (6)"},{"comment":"The validation is circular. The polynomial coefficients describing ⟨E_spec⟩, ⟨N_hits⟩, their variances, and their covariance are obtained from the same DCM-QGSM-SMM sample later used to validate the reconstruction; the parameters α_E, α_N, β_E, β_N, and ε are also fitted to that same two-dimensional distribution. The text itself states that α_E and α_N are close to 1 and β_E, β_N close to zero because the fit function is based on the same simulation. Thus the <2% agreement in Fig. 10 demonstrates consistency of the parameterization with the sample, not predictive accuracy. To support the claim, the authors need an out-of-sample test (e.g., fit on one subset and validate on another), a closure test with injected parameters, or a comparison with a standard centrality method.","section":"§V, Eqs. (8)–(11), Fig. 10"},{"comment":"The derivation leading to the β_N parameter appears internally inconsistent. If one sets n=α_N n_MC, then σ_n^2=α_N^2 σ_{n_MC}^2, so the combination σ_n^2−σ_{n_MC}^2 α_N^2 in Eq. (16) vanishes identically and β_N=0. The text appears to treat σ_n^2 and σ_{n_MC}^2 as independent while simultaneously imposing a linear scaling. This undermines the parameterization in Eqs. (10)–(11). The assumptions behind the scaling should be clarified or the derivation corrected.","section":"Appendix, Eqs. (14)–(17)"}],"minor_comments":[{"comment":"The integral over N_hits is written with lower and upper limits both equal to N1; it should presumably be from N1 to N2. Please correct this typo.","section":"§V, Eq. (12)"},{"comment":"Several typos appear: 'fitst physics run' (should be 'first'), 'dependenc'/'varianc' in Fig. 4, and 'deposed energy' in Fig. 3. These are presentation issues but should be fixed.","section":"§III and figure captions"},{"comment":"The text calls Eq. (6) a two-dimensional Gamma distribution, but it is a product of two independent Gamma distributions. If a true bivariate Gamma is intended, the relation should be clarified; otherwise the terminology is misleading.","section":"§IV, terminology"}],"recommendation":"major_revision","confidential_remarks":"For the editor: the missing factor of 1/2 in the rotation angle is localized but load-bearing. I believe the paper can be repaired by correcting the angle, justifying or empirically testing the factorized-Gamma kernel, and adding non-circular validation. If the authors cannot supply these, rejection would be warranted. I therefore recommend major revision rather than outright rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper's real contribution is a two-dimensional Bayesian centrality estimator that replaces the usual Gaussian kernel with a product of two Gamma distributions after a rotation of the two observables. For a low-energy fixed-target experiment like BM@N, where a single observable (spectator energy) is non-monotonic in impact parameter, the two-observable approach is well motivated, and the idea of using Gamma marginals is sensible for skewed distributions in the peripheral region. The appendix's derivation of the mean-variance scaling from Glauber fluctuations is a thoughtful addition and reduces free parameters.\n\nThe problem is the load-bearing math. Section IV states that the rotation angle phi = arctan(2Cov/(sigma_E^2 - sigma_N^2)) makes the covariance zero. It does not; diagonalizing a 2x2 covariance requires phi = (1/2) arctan(2Cov/(sigma_E^2 - sigma_N^2)). With their angle, the off-diagonal element is not zero, so the factorization P(Espec,Nhits|b) = Gamma * Gamma in Eq. (6) does not follow. Even with the corrected angle, the independence and Gamma-marginal assumptions are modeling choices, not derived properties. This is not a cosmetic issue: Eq. (6) is the kernel that generates all the impact-parameter reconstructions, so the claimed 2% agreement in Fig. 10 is supported only if the implementation silently uses a different angle than the text provides.\n\nThe validation is also weak. The mean, variance, and covariance parameterizations are fitted to the same DCM-QGSM-SMM sample used for the validation, the fit parameters come out near unity, and there are no baselines - a 2D Gaussian or the 1D Bayesian method would tell you whether the extra complexity buys anything. The 2% agreement is a consistency check, not an independent test.\n\nAll that said, the idea is not dead. The half-angle fix is trivial, and the Gamma product ansatz could be tested against real BM@N data or an independent event generator. As written, the central derivation is wrong, so I would not accept it for publication. But it deserves a serious referee: the problem is real, the approach is novel in this application, and a competent referee can point the authors to a repairable path. Bring it to the reading group as a case study in how a factor-of-two error can undermine a paper's main claim.","headline":"A two-observable Gamma kernel for centrality at BM@N is a good idea, but the rotation formula is wrong and the validation is in-sample, so the paper needs major revision.","tokens_in":9230,"tokens_out":3657,"would_cite":false,"duration_ms":36529,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["25.75.-q"],"model":"deepseek-v4-flash","headline":"This paper claims that a two-dimensional Bayesian centrality method, built on a product of two Gamma distributions for the joint fluctuations of track-hit count and forward spectator energy, reconstructs the impact-parameter distribution in","keywords":["centrality determination","impact parameter","Bayesian reconstruction","Gamma distribution","heavy-ion collisions","spectator energy","track multiplicity","two-dimensional observable analysis"],"falsifier":"Take the same simulated events, compute the two rotated variables X1 and X2 using the paper's angle phi = arctan(2Cov/(sigma_E^2 - sigma_N^2)), and evaluate their sample covariance: if it is substantially different from zero, the product-of-Gamma kernel does not describe the true joint distribution. A complementary check is to replace phi by (1/2)arctan(2Cov/(sigma_E^2 - sigma_N^2)) and re-fit; if the reconstructed mean-impact-parameter curve shifts by more than a few percent, the 2% agreement is tied to the rotation convention rather than to the physics.","tokens_in":8275,"feed_emoji":"⚛️","tokens_out":7826,"duration_ms":73856,"temperature":0.7,"pith_summary":"The paper proposes that centrality in nucleus-nucleus collisions can be determined more reliably by using two correlated observables at once—the number of charged-particle track hits and the energy deposited by spectator nucleons in a forward calorimeter—instead of a single observable. Its central claim is that the joint fluctuation of these two quantities at a fixed impact parameter is well described by the product of two Gamma distributions, obtained after a rotation that is intended to decorrelate the observables, and that this kernel yields an accurate Bayesian reconstruction of the impact-parameter distribution. A sympathetic reader would care because at the low beam energies and small systems considered here, single-observable centrality estimators are skewed in both central and peripheral regions, and a faithful reconstruction of the impact parameter is what makes comparisons with theoretical models meaningful. The quantitative result is that, in a realistic simulation of Xe+CsI collisions at 3.8A GeV, the reconstructed mean impact parameter as a function of centrality agrees with the true generator values within 2%.","feed_headline":"Bayes with two observables recovers impact parameter within 2%","feed_subtitle":"Paired Gamma-distributed signals sharpen impact-parameter reconstruction in heavy-ion collisions.","key_machinery":"Equation (6), P(N_hits, E_spec|b) = Gamma(k1, theta1) * Gamma(k2, theta2), is the load-bearing object: a factorized kernel built from the two rotated observables. The rotation is introduced to decorrelate the two measurements, the means and variances of the rotated variables come from Eqs. (2)–(5), and the Gamma parameters are derived from those means and variances. Bayes' theorem (Eq. (12)) then turns a measured window of the two observables into a posterior probability over the impact parameter, and constrained k-means clustering divides the two-dimensional distribution into centrality classes. The paper also derives a decomposition of multiplicity fluctuations into an ancestor-number comp","core_discovery":"In the paper's own terms, the discovery is an ansatz for the conditional probability of two detector signals at fixed impact parameter: after a rotation of the track-hit count and spectator energy into new variables with vanishing covariance, each is modeled by a Gamma distribution, so the joint density is the product of two Gammas. The shape and scale parameters are fixed by the impact-parameter-dependent means and variances through k = mean^2/variance and theta = variance/mean, and the whole two-dimensional density is fitted to simulated detector output with a small set of scaling parameters that absorb differences between data and simulation. The fit reproduces the two-dimensional distrib","pith_inferences":["The rotation angle in Section IV is not the standard diagonalizing angle: the covariance of the rotated variables is not actually zero with the stated formula, so the factorization into independent Gammas is an additional model assumption rather than a consequence of the rotation. A corrected angle or a direct numerical check of Cov(X1,X2)=0 would settle this internally.","If the product-of-Gamma assumption holds for this pair of observables, the same scheme could in principle be extended to three or more observables, although independence of more than two rotated Gamma variables is unlikely to survive without a more general copula.","The validation is done on simulations where the fit function is built from the same generator as the data; a stronger test would apply the fit to an independent event generator or to real data with the calibration parameters left free.","A direct, testable signature of the paper's assumption is that conditional slices of the two-dimensional distribution at fixed impact parameter should be Gamma-shaped; event generators that store the true impact parameter could verify this slice-by-slice."],"forward_implications":["If the kernel is right, centrality classes can be defined from two observables simultaneously, reducing the autocorrelation that arises when the same multiplicity used for classification is also used for physics analyses.","The method provides a parameterized bridge between simulations and data: the alpha, beta, and epsilon parameters calibrate means, variances, and trigger losses, so the approach can be applied to real detector runs after tuning.","The Gamma-based kernel extends reliable centrality reconstruction into peripheral and very central windows where skewness makes a normal kernel inadequate.","The 2% agreement of the mean impact parameter versus centrality, if it survives in real data, would permit quantitative comparisons of centrality-dependent observables across models and experiments."],"fun_headline_variants":["Bayesian 2D fit of track hits and spectator energy boosts centrality","Impact parameter from two observables: a Bayesian twist","Covariance-removed Gammas sharpen centrality in Xe+CsI","Two Gamma signals pin down impact parameter in heavy-ion collisions","Bayesian centrality with paired Gamma distributions improves precision"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole construction depends on the assumption that a certain rotation of the two measured signals produces two independent, Gamma-distributed variables, so that their joint probability factorizes exactly as in Eq. (6); if the rotation does not actually eliminate the correlation, or the signals are not Gamma-shaped, the reconstructed impact-parameter distributions inherit a bias.","fun_headline_variants_meta":{"raw":{"variants":["Bayesian 2D fit of track hits and spectator energy boosts centrality","Impact parameter from two observables: a Bayesian twist","Covariance-removed Gammas sharpen centrality in Xe+CsI","Two Gamma signals pin down impact parameter in heavy-ion collisions","Bayesian centrality with paired Gamma distributions improves precision"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000136,"raw_usage":{"total_tokens":935,"prompt_tokens":646,"completion_tokens":289,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":390,"completion_tokens_details":{"reasoning_tokens":202}},"tokens_in":390,"tokens_out":289,"duration_ms":3414,"temperature":1.0,"reasoning_tokens":202,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T07:14:05.833981+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the same simulated events, compute the two rotated variables X1 and X2 using the paper's angle phi = arctan(2Cov/(sigma_E^2 - sigma_N^2)), and evaluate their sample covariance: if it is substantially different from zero, the product-of-Gamma kernel does not describe the true joint distribution. A complementary check is to replace phi by (1/2)arctan(2Cov/(sigma_E^2 - sigma_N^2)) and re-fit; if the reconstructed mean-impact-parameter curve shifts by more than a few percent, the 2% agreement is tied to the rotation convention rather than to the physics.","supporting_citations":[],"review_version":1}