{"id":"085b62de-aada-4f1b-935f-01899edbadb0","arxiv_id":"2506.20633","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"rd-spiral provides a transparent pseudo-spectral simulator for 2D reaction-diffusion spirals and claims characteristic statistical differences between stable, turbulent, and decaying regimes.","lead":"This paper introduces rd-spiral, an open-source Python library that simulates 2D reaction-diffusion spiral waves using FFT-based spectral methods and adaptive time stepping. It reports statistical signatures that distinguish stable spirals from turbulent chaos, including very high kurtosis and changes in mutual information between activator and inhibitor fields.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Regime-comparison statistics are built on autocorrelated single-run time series, and the stable-series 'kurtosis > 96' reflects a near-constant sigma_u record rather than stable-spiral dynamics.","rationale":"The reader's weakest assumption — that the time points are treated as independent samples — captures the core of the problem, and I agree that this invalidates the reported p-values, confidence intervals, and effect sizes. I extend this concern in two ways. First, the stable sigma_u series is so narrowly concentrated (CV ~0.0026) that the excess kurtosis of 96 is most plausibly an artifact of a near-point-mass distribution plus a few transient points, rather than a meaningful characterization of stable spiral dynamics. Second, the mutual-information decline numbers in the abstract and conclusions do not match the values reported in Section 3.2; at t=5 and t=100 the stable decline is about 6.1% and the turbulent decline about 20.4%, so the stated 6.5% and 10.7% are not traceable. Neither issue requires rejecting the software contribution: the code and archived data are real, and the qualitative observation that stable spirals and turbulent fragmentation differ is credible. But the paper's headline quantitative claims need reanalysis with effective-sample-size-aware or ensemble-based statistics before they can be accepted. Because those claims are fixable and the library itself is plausibly usable, the conditional verdict stands.","tokens_in":28280,"tokens_out":6904,"duration_ms":85726,"concrete_test":"Using the archived OSF outputs, estimate the integrated autocorrelation time tau of each sigma_u and sigma_v series and compute the effective sample size N_eff = N / (1 + 2 sum_{k>=1} rho_k). Then redo the stable-vs-turbulent Mann-Whitney/KS/Cliff's-delta comparisons with a block bootstrap using block length >= 5 tau, and recompute excess kurtosis on the stable sigma_u series using only the post-transient window t in [100,200]. If N_eff drops below approximately 30, or the post-transient kurtosis falls below 10, or the block-bootstrap p-values become non-significant, the reported extreme statistics and effect sizes are artifacts of autocorrelation and transient contamination rather than robust regime differences.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative claims that carry the abstract and Section 3.2 — excess kurtosis > 96, Mann-Whitney/KS p < 0.001, Cliff's delta 0.37–0.78, and the 6.5% vs 10.7% coupling decline — all depend on treating every time sample of sigma_u(t) and sigma_v(t) as an independent draw. Section 2.3.1 constructs these series from single runs, and the two regimes differ in grid size (128 vs 256), domain size (L=20 vs 50), integration time (t=200 vs 500), and initial charge (m=1 vs 4). Stable sigma_u is essentially constant: mean 0.670465, SD 0.001711, CV 0.0026, IQR 0.000255. For such a nearly constant deterministic series, excess kurtosis 96 is a point-mass artifact, not a dynamical signature; normality tests reject Gaussianity trivially. The iid assumption underlying Mann-Whitney, KS, bootstrap CIs, and Cliff's delta is false, so the nominal sample sizes of 2001 and 5001 overstate the information content by orders of magnitude. In addition, the reported mutual-information declines are not reproducible from the values given in Section 3.2: from t=5 to t=100, stable I(U;V) falls from 4.5867 to 4.3048 bits (−6.1%) and turbulent I(U;V) falls from 4.6397 to 3.6911 bits (−20.4%), not the stated −6.5% and −10.7%. The open-source library itself may well be functional, but the central statistical comparisons are unsupported without block/ensemble inference or multiple seeded runs.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces rd-spiral, an open-source Python library for simulating two-dimensional reaction-diffusion systems using a pseudo-spectral spatial discretization combined with adaptive Dormand-Prince time integration. The authors present three parameter regimes—stable spiral rotation, spatiotemporal chaos, and pattern decay—and use the resulting simulations to make quantitative claims: excess kurtosis above 96 in stable dynamics, non-parametric effect sizes (Cliff's delta 0.37–0.78), and a mutual-information decline of 6.5% in stable versus 10.7% in turbulent regimes. The manuscript also documents software features such as checkpointing, automated equilibrium classification, and archived simulation outputs.","tokens_in":28737,"tokens_out":5575,"duration_ms":66176,"significance":"If the statistical claims were fully supported, the paper would offer a useful pedagogical and research tool plus a quantitative signature separating stable and turbulent spiral regimes. The manuscript has real strengths: the solver design is standard and transparent, the code is open source, the simulation outputs are archived on OSF, and the pedagogical emphasis on code clarity is commendable. However, the headline quantitative claims are currently not supported: the regime comparisons are built on autocorrelated single-run time series, the kurtosis result is an artifact of a nearly constant scalar series, and the reported mutual-information declines are not reproducible from the numbers given in the text. These issues are load-bearing for the abstract and conclusions, so the manuscript needs substantial revision before publication.","major_comments":[{"comment":"The regime-comparison statistics are computed from single simulations with different numerical setups: the stable spiral uses a 128×128 grid, L=20, t=0–200, and m=1, while the turbulent case uses 256×256, L=50, t=0–500, and m=4. The σu(t) and σv(t) series are strongly autocorrelated—the stable σu series is nearly constant (mean 0.670465, SD 0.001711, CV≈0.0026)—so treating each time sample as an independent observation invalidates the Mann-Whitney U, Kolmogorov-Smirnov, bootstrap confidence interval, and Cliff's delta calculations reported in Section 3.2. The p<0.001 values and δ=0.37–0.78 therefore do not support the claimed quantitative regime distinctions; block bootstrap or ensemble inference over multiple independent initial conditions with matched grids and domains is required.","section":"Section 2.3.1 and Section 3.2"},{"comment":"The claim of 'extreme departures from Gaussian behavior in stable spiral dynamics' rests on excess kurtosis values of 96.2497 and 96.4037 for σu and σv. The stable σu series has mean 0.670465, standard deviation 0.001711, and IQR 0.000255, so its distribution is effectively a point mass. The large kurtosis and the normality-test rejections are dominated by tiny numerical fluctuations rather than by spiral-wave dynamics, making this statistic an artifact rather than a meaningful dynamical signature. Please replace this analysis with characterizations of the field distributions or of variability across independent realizations, or clearly relabel the statistic as a property of the scalar monitoring series.","section":"Section 3.2, Eqs. (58)–(59)"},{"comment":"The stated mutual-information declines are not reproducible from the values in the text. From t=5 to t=100, the stable I(U;V) changes from 4.5867 to 4.3048 bits (−6.1%), while the turbulent I(U;V) changes from 4.6397 to 3.6911 bits (−20.4%); neither matches the claimed −6.5% and −10.7%. If the intended comparison is between t=5 and t=200, the final I(U;V) values are not reported. Please correct the percentages and provide a complete table of H(U), H(V), H(U,V), and I(U;V) at all reported times, or remove the quantitative comparison from the Abstract.","section":"Section 3.2 (temporal evolution summary), Abstract, and Conclusions"}],"minor_comments":[{"comment":"The derivation claims to proceed from conservation laws, but the reduction to the canonical form relies on specific modeling choices—truncation to odd cubic terms, equal linear growth rates, symmetric saturation, and a single coupling β. Please state explicitly that these are assumptions rather than consequences of the conservation-law derivation, and cite the prior source of the canonical model more prominently.","section":"Section 2.1, Eqs. (34)–(35)"},{"comment":"The paper asserts exponential convergence and spectral accuracy, but no numerical convergence experiment (e.g., a manufactured solution or comparison with a known exact solution) is reported. A short convergence study would substantiate this central numerical claim.","section":"Section 2.2, 'Error analysis' paragraph"},{"comment":"Resampling the original autocorrelated σu and σv series with replacement does not validate the normality-test results, because the bootstrap samples inherit the same non-independence. This point is already implicit in the first major comment, but the text should not describe the bootstrap procedure as providing validation of the reported p-values.","section":"Section 2.3.4, Bootstrap for normality tests"}],"recommendation":"major_revision","confidential_remarks":"The software archive and open-source code appear to be potentially useful contributions, and I did not find evidence of input-output circularity in the sense that the regime parameters are hand-chosen rather than fitted to the simulation outputs. The main obstacles are statistical: the pseudoreplication in the regime comparisons and the incorrect mutual-information arithmetic. I would be willing to review a revised version that replaces the single-run comparisons with matched multi-run or block-based inference and corrects the quantitative summary statements. The Introduction's self-citations on Python performance are peripheral and should be trimmed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on arXiv:2506.20633. The genuine contribution is the open-source library rd-spiral: MIT-licensed, archived code and data, clear pseudo-spectral implementation (FFT + Dormand-Prince) with checkpointing and equilibrium classification. That part is solid and useful for teaching. The paper reproduces three qualitative regimes (stable spiral, turbulent, decay) which is exactly what an educational tool should do.\n\nThe soft spots are in the statistics. The headline claims—excess kurtosis > 96, Mann-Whitney/KS p-values, Cliff's delta 0.37–0.78, and the 6.5% vs 10.7% mutual-information decline—are not supported. The time series sigma_u(t) and sigma_v(t) are treated as independent samples, but they come from single runs with different grid sizes (128 vs 256), domains (L=20 vs 50), integration times, and initial charges. The series are also strongly autocorrelated. So the p-values and CIs overstate the information. The stable sigma_u is essentially constant (mean 0.670, SD 0.0017), so an excess kurtosis of 96 is a point-mass artifact, not a dynamical signature. And the reported mutual-information declines don't match the numbers in Section 3.2: 4.5867 to 4.3048 bits is −6.1%, and 4.6397 to 3.6911 is −20.4%, not the stated −6.5% and −10.7%. That is an easy check and it's wrong.\n\nThe qualitative distinction between regimes is likely real—the images show it—so the core story is fine. But the quantitative statistical framework needs to be redone: multiple seeded runs, block bootstrap or ensemble statistics, and corrected MI percentages. The \"derivation from first principles\" is also overblown; the model is the known Bramburger system, and the Taylor-expansion path to it is a pedagogical re-derivation, not a new result.\n\nWho is this for? Educators and students who want a transparent, working 2D reaction-diffusion solver. It is not a research contribution beyond the software itself. With the statistics fixed and the claims scaled down, it could be a solid software/education paper.\n\nMy recommendation: send it to peer review, but as a software/education manuscript with major revisions required. The library deserves referee time; the current quantitative claims do not.","headline":"A transparent educational RD solver, but the headline statistics are artifacts of pseudoreplication and a near-constant time series.","tokens_in":29205,"tokens_out":3322,"would_cite":false,"duration_ms":35041,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K57","65M70","62G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that its open-source pseudo-spectral solver rd-spiral captures all three regimes of 2D spiral-wave dynamics—stable rotation, spatiotemporal chaos, and pattern decay—and that the simulations reveal extreme non-Gaussian…","keywords":["reaction-diffusion systems","pseudo-spectral methods","spiral waves","nonlinear dynamics","pattern formation","spatiotemporal chaos","non-parametric statistics","information theory"],"falsifier":"Compute the integral autocorrelation time of $\\sigma_u(t)$ in the stable run; if it is comparable to the total integration time, the effective independent sample size is tiny, and the reported p-values would not survive. Equivalently, run the stable configuration at the turbulent run's resolution and domain ($256\\times256$, $L=50$, $t=500$) and repeat all statistics: if excess kurtosis drops below a few units or the mutual-information decline becomes comparable across regimes, the claimed signatures are artifacts of the single-run setup.","tokens_in":28107,"feed_emoji":"🌀","tokens_out":10930,"duration_ms":101286,"temperature":0.7,"pith_summary":"rd-spiral is a new open-source Python library for simulating two-dimensional reaction-diffusion systems with a pseudo-spectral method: fast Fourier transforms handle spatial derivatives, and an adaptive Runge–Kutta integrator advances time. The paper sets out to show that this single solver can reproduce the full spectrum of spiral-wave behavior—stable rotating spirals, spatiotemporal turbulence, and complete pattern decay—with simple parameter changes. The authors then use the simulated fields to argue that stable spiral dynamics have strongly non-Gaussian amplitude statistics, with excess kurtosis greater than 96, while turbulent dynamics are much closer to Gaussian. They also find that mutual information between the activator and inhibitor fields declines more sharply in turbulence (10.7%) than in stable regimes (6.5%), which they read as a quantitative marker of pattern breakup. If these claims hold, the library is both a transparent teaching tool and a validated numerical platform for studying pattern formation.","feed_headline":"Stable spiral waves show extreme non-Gaussian statistics","feed_subtitle":"A new pseudo-spectral Python library simulates stable, chaotic, and decaying spirals; excess kurtosis exceeds 96.","key_machinery":"The load-bearing piece is the pseudo-spectral spatial discretization: the periodic domain admits a Fourier expansion, so the Laplacian becomes multiplication by $-k^2$ in spectral space, giving exponential convergence for smooth fields, while the cubic reaction terms are evaluated pointwise in physical space to avoid costly convolutions. The resulting semi-discrete ODE system is integrated by the adaptive Dormand–Prince RK5(4) scheme, which the paper reports handles stiffness ratios above $6{:}1$. The statistical claims rest on the time series of the spatial standard deviations $\\sigma_u(t)$ and $\\sigma_v(t)$, which are analyzed with normality tests, bootstrap resampling, Cliff's delta, and entropy and mutual information.","core_discovery":"The central claim is that rd-spiral successfully captures the complete dynamical spectrum of the two-species reaction-diffusion system $\\partial_t u = D_1 \\nabla^2 u + u - u^3 - uv^2 + \\beta(u^2v+v^3)$ and its counterpart for $v$, across three regimes: stable spiral rotation ($D_1=D_2=0.1$, $\\beta=1.0$), spatiotemporal chaos ($D_1=0.03$, $D_2=0.20$, $\\beta=0.65$), and pattern decay ($D_1=D_2=0.5$). Analyzing the time series of the spatial standard deviations $\\sigma_u(t)$ and $\\sigma_v(t)$, the authors report extreme leptokurtosis in the stable regime (excess kurtosis $>96$), unanimous rejection of normality across all four tests in all bootstrap samples, and effect sizes of Cliff's $\\delta = 0.37$ (for $\\sigma_u$) and $\\delta = 0.78$ (for $\\sigma_v$) that separate the regimes. Information-theoretic analysis shows a $10.7\\%$ reduction in activator-inhibitor mutual information during turbulent fragmentation, versus $6.5\\%$ in stable spirals, which the paper interprets as a quantitative measure of field decoupling during spiral breakup.","pith_inferences":["Editorial extension: because the reported statistics come from single long runs rather than ensembles, the same protocol could be applied to many random initial phases to obtain distribution-based error bars on kurtosis and mutual-information decline.","Editorial extension: the mutual-information decline of the turbulent regime suggests a possible early-warning indicator for spiral breakup that could be monitored in cardiac or chemical excitable media, not just in this model.","Editorial extension: the analysis pipeline (standard-deviation time series, normality tests, bootstrap, Cliff's delta, mutual information) transfers directly to other two-variable excitable-media models to test whether the reported non-Gaussian and decoupling signatures are generic.","Editorial extension: replacing the fixed grid with adaptive spatial resolution could extend the library to regimes where the spiral core develops extremely steep gradients, where spectral accuracy on a uniform grid degrades."],"forward_implications":["A single open-source solver, with three documented parameter sets, now covers the three qualitative behaviors of reaction-diffusion spirals, making it a ready-made platform for teaching nonlinear dynamics.","Stable spiral amplitude statistics are strongly non-Gaussian; any parametric analysis that assumes Gaussianity for such regimes is likely to be invalid.","Cliff's delta of $0.37$ and $0.78$ for $u$ and $v$, together with mutual-information declines of $6.5\\%$ and $10.7\\%$, provide quantitative separators for classifying a spiral-wave state as stable or turbulent.","The method tolerates diffusion-coefficient ratios over $6{:}1$, indicating the pseudo-spectral plus adaptive-integrator approach can handle stiff pattern-forming systems without specialized exponential integrators.","Checkpointing and the preserved initial conditions allow long turbulent simulations to be restarted and reproduced exactly, which supports reproducibility claims for extended runs."],"supporting_citations":[{"why":"Supplies the dimensionless two-species reaction-diffusion system (Eqs. 34–35) that the solver integrates.","marker":"[7]"},{"why":"Provides the spectral differentiation framework that justifies the FFT-based Laplacian.","marker":"[61]"},{"why":"Supplies the Dormand–Prince embedded RK5(4) pair used for adaptive time integration.","marker":"[17]"},{"why":"Motivates the pseudo-spectral evaluation of nonlinear terms in physical space.","marker":"[47]"},{"why":"Defines Cliff's delta, the effect-size statistic used to separate the regimes.","marker":"[10]"},{"why":"Supplies the entropy and mutual-information measures used for coupling analysis.","marker":"[55]"},{"why":"Provides the bootstrap resampling method behind the confidence intervals and normality-test stability checks.","marker":"[18]"},{"why":"Supplies the ODE solver and statistical testing routines used in the implementation.","marker":"[63]"},{"why":"Defines the Shapiro–Wilk test central to the normality-rejection claims.","marker":"[56]"},{"why":"Defines the Jarque–Bera test used for the kurtosis-based normality assessment.","marker":"[32]"}],"fun_headline_variants":["Stable spiral kurtosis tops 96","Turbulence cuts spiral coupling 10.7%","Extreme kurtosis in stable spirals","Non-Gaussian spiral waves simulated"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The statistical comparisons treat the consecutive values of $\\sigma_u(t)$ and $\\sigma_v(t)$ within each run as independent samples, although they come from single, strongly autocorrelated trajectories that also differ in grid size, domain size, integration time, and initial vortex charge between the stable and turbulent runs.","fun_headline_variants_meta":{"raw":{"variants":["Stable spiral kurtosis tops 96","Turbulence cuts spiral coupling 10.7%","Extreme kurtosis in stable spirals","Non-Gaussian spiral waves simulated"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001312,"raw_usage":{"total_tokens":5372,"prompt_tokens":995,"completion_tokens":4377,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":611,"completion_tokens_details":{"reasoning_tokens":4319}},"tokens_in":611,"tokens_out":4377,"duration_ms":39314,"temperature":1.0,"reasoning_tokens":4319,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:44:17.750110+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the integral autocorrelation time of $\\sigma_u(t)$ in the stable run; if it is comparable to the total integration time, the effective independent sample size is tiny, and the reported p-values would not survive. Equivalently, run the stable configuration at the turbulent run's resolution and domain ($256\\times256$, $L=50$, $t=500$) and repeat all statistics: if excess kurtosis drops below a few units or the mutual-information decline becomes comparable across regimes, the claimed signatures are artifacts of the single-run setup.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Motivates the pseudo-spectral evaluation of nonlinear terms in physical space."}],"review_version":1}