{"id":"ca196920-5710-4b85-b4f3-363d842225c6","arxiv_id":"2507.06255","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"Betti numbers, Euler characteristic and their sum for excursion sets are expressed in terms of Binomial coefficients of a topological basis, predicting Gaussian statistics except at high thresholds.","lead":"The paper represents the connected pieces and holes of random fields' excursion sets as counts of simple topological building blocks, and models those counts with Binomial distributions. It argues this explains when Betti numbers and the Euler characteristic are approximately Gaussian, which matters for how cosmologists estimate errors.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's own negative covariance between b0 and b1 contradicts the independence assumption on which the Binomial-product derivation rests.","rationale":"The reader's REJECT verdict is justified, and this stress-test strengthens it. The load-bearing weak point is indeed the independence/Binomial assumption on the coefficients m_j. The reader framed it as an unproved simplifying assumption; the stronger problem is that the paper's own Fig. 4 reports a negative covariance between b0 and b1, which is algebraically impossible for independent nonnegative m_j. This directly invalidates the factorized joint PDF in Eq. (19) and the product PDF in Eq. (44), and with them the Binomial-sum derivations and the Gaussian conclusions. The multinomial interpretation of the colored-ball model is also internally inconsistent with the independence assumption. A direct decomposition of the simulations into basis coefficients would settle the issue, but the burden of proof is on the model, and the reported numerics already point against it. I therefore keep the reader's verdict unchanged.","tokens_in":17616,"tokens_out":10080,"duration_ms":124770,"concrete_test":"Use the same 10^4 Gaussian simulations as in Fig. 4. At a fixed threshold ν, decompose every excursion set into connected components and record (m_0, m_1, ...). Then (i) estimate Cov(b0,b1) and check whether it is nonnegative, as independence requires; (ii) test the product structure P(m0,m1) = P(m0)P(m1) with a chi-square or mutual-information test; (iii) check whether N_tot = Σ_j m_j is constant across realizations, as the N-trial multinomial model requires, or variable, as independent Binomials require. If the empirical covariance is negative or the product test fails, Eq. (44) is rejected and the central claim does not follow.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central distributional claim depends on the coefficients m_j(ν) being independent at each threshold: Eq. (19) factorizes their joint PDF, and Eq. (44) writes P(b0) as a product of marginal Binomial PDFs. But Sec. IV.B.1 reports Cov(b0,b1) < 0 (Fig. 4 and Eq. 34). If the m_j are independent nonnegative variables, then Cov(b0,b1) = Σ_{j≥1} j Var(m_j) ≥ 0, so a negative covariance is impossible under the independence assumption. The colored-ball model in Sec. IV.B.3 also points the other way: N draws from a box of colored balls yield multinomial counts with Cov(m_j,m_k) = -N p_j p_k, so the coefficients are negatively correlated, not independent. The numerical calibration of N and p_j from the same ensemble mean and variance cannot repair this, because it imposes only the first two moments of each statistic and does not test the product structure. Hence the derived PDFs for b0, b1, χ, and bsum, and the CLT conclusion, do not follow from the model as stated; the independence assumption is not merely unproved but contradicted by the paper's own numerical results.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a representation of the excursion sets of smooth random fields as finite unions of a topological basis, with nonnegative integer coefficients m_j counting the numbers of connected components of each homological type. Betti numbers b0 and b1, the Euler characteristic χ, and the sum of Betti numbers bsum are then expressed as linear combinations of these coefficients. The paper assumes that the m_j are statistically independent and models each as a Binomial variable, and it uses this to derive that the topological statistics are asymptotically Gaussian when the packing parameter q=(L/r_c)^d tends to infinity. The claims are tested numerically on Gaussian CMB maps at several smoothing scales, and the paper identifies threshold regimes where Gaussian approximations hold. The central assertion is that the Binomial nature of the coefficients controls the statistical nature of the topological statistics.","tokens_in":17841,"tokens_out":5987,"duration_ms":66147,"significance":"If the distributional claims were sound, the paper would provide a useful framework for attaching error bars to Betti numbers and Euler characteristics in cosmological analyses, and the observation that bsum has smaller variance than b0 or b1 separately is potentially valuable. The topological-basis representation itself is a transparent bookkeeping device, and the numerical experiments are extensive and clearly described. However, the paper provides no derivation of the Binomial model, its own numerical results contradict the independence assumption on which the model rests, and the numerical validation is circular because the model parameters are calibrated from the very quantities used for comparison. The central probabilistic content is therefore not established, and the asymptotic Gaussian conclusion is unsupported.","major_comments":[{"comment":"The independence assumption stated in Sec. IV.B is contradicted by the paper's own numerical results. Since b0 = Σ m_j and b1 = Σ j m_j, if the m_j are independent then Cov(b0,b1) = Σ_{j≥1} j Var(m_j) ≥ 0. The paper reports Cov(b0,b1) < 0 from the simulations (Fig. 4, Eq. (34)). This makes the factorization in Eq. (19) and the product formula in Eq. (44) internally inconsistent with the model used to interpret the numerical data. The discussion following Eq. (34) attempts to reinterpret the inequality through the variances, but it does not repair the sign contradiction with independence.","section":"Sec. IV.B and Sec. IV.B.1, Eqs. (19), (30), (34)"},{"comment":"The colored-ball model of Sec. IV.B.3 describes drawing N balls from a box with distinct colors, which yields multinomial counts with Cov(m_j,m_k) = -N p_j p_k < 0 for j ≠ k. The coefficients are therefore negatively dependent, not independent, under the very analogy used to justify the Binomial assumption. Moreover, the note in Sec. IV.B.4 that Binomial modeling corresponds to 'Bernoulli trials without replacement' is erroneous: Binomial counts arise from independent trials (sampling with replacement, or infinite population), while sampling without replacement gives hypergeometric counts. The statistical model is thus internally inconsistent.","section":"Sec. IV.B.3 and Sec. IV.B.4"},{"comment":"The numerical validation is circular. The values of N(ν) and p_j(ν) are fixed by equating the Binomial formulas for the mean and variance to the analytic mean of χ and the numerically computed standard deviation σ_χ (or, in the intermediate regime, to the ensemble mean and standard deviation of each statistic). The resulting Binomial PDFs are then compared to the numerical PDFs obtained from the same ensemble. A two-parameter family will generally match the first two moments of a given distribution; this procedure does not test whether the m_j are Binomial, does not test independence, and does not test the product structure of Eq. (44). The agreement displayed in Fig. 5 is therefore not evidence for the central model.","section":"Sec. IV.B.4, Eqs. (55)-(56), (61)-(62)"},{"comment":"The central limit theorem argument is not a proof. It invokes block-sum independence without specifying the dependence structure or verifying a mixing condition for the m_j's, and it does not verify that the variance of the sum diverges. Boundedness of the summands is not sufficient for asymptotic normality; a Lindeberg-type condition is required. More fundamentally, since the m_j are dependent for the Gaussian fields studied in the paper (as shown by the negative covariance in Sec. IV.B.1), the classical CLT does not apply to b0, b1, χ, or bsum. The conclusion that these statistics 'tend to Gaussian random variables provided q → ∞' is therefore unsupported.","section":"Sec. IV.B.2"}],"minor_comments":[{"comment":"There is a typo: 'distinguishuable' should be 'distinguishable'.","section":"Sec. III"},{"comment":"The notation 'N C_k p^k (1-p)^{N-k}' is nonstandard; the binomial coefficient is usually written '\\binom{N}{k}' or 'C(N,k)'.","section":"Sec. IV.B.3"},{"comment":"The statement that 'b0(ν) = b1(−ν)' as a reflection of Alexander duality is an approximate identity for finite maps and should be qualified as approximate; it is not an exact equality for a single finite-volume realization.","section":"Sec. IV.B.4"},{"comment":"Reference [2] is listed with a title and URL identical to Reference [1]; the correct reference is R. J. Adler and J. E. Taylor, 'Random Fields and Geometry' (Springer, 2007).","section":"References"},{"comment":"The phrase 'we are assuming Bernoulli trials without replacement' is not only incorrect but also unconceptual; consider rewriting the modeling discussion to distinguish independent trials (Binominal) from finite-population sampling (hypergeometric).","section":"Sec. IV.B.4"},{"comment":"The caption states that 'the values of N, p shown are obtained using the ensemble expectation and standard deviation of each statistic, at each ν, as inputs in the Binomial distribution.' This explicitly confirms the circular calibration procedure and should be flagged as a limitation of the validation.","section":"Fig. 5 caption"}],"recommendation":"reject","confidential_remarks":"The contradiction between the assumed independence of the m_j and the paper's own negative covariance estimate for (b0,b1) is decisive. Even if the topological-basis representation is a useful reformulation, the distributional conclusions do not follow from the stated assumptions, and the circular validation cannot rescue them. The author may wish to develop a model that incorporates the negative correlations (e.g., a multinomial or another dependent model) and to test joint predictions that are not used in the calibration. As it stands, the manuscript does not meet the standard for publication in a mathematical statistics journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is easy to summarize: excursion sets are written as unions of basic pieces Q_j (disks with j holes), with nonnegative integer coefficients m_j. Betti numbers, Euler characteristic, and bsum are linear combinations of the m_j's, and the author proposes to model the m_j's as independent Binomial variables so that the topological statistics inherit their distributions as sums of Binomials. There is a real kernel of value here: the representation is a clean bookkeeping device, the generating function h(α) is neat, and the numerical study of 10^4 CMB maps is solid work. The practical observation that bsum has smaller variance than b0 or b1 because b0 and b1 are anti-correlated is useful and likely robust.\n\nThe soft spots are concentrated in the probabilistic content. The independence assumption on the m_j's is not just unproved; it is contradicted by the paper's own numbers. Under independence, Cov(b0,b1) = Σ j Var(m_j) ≥ 0, but Fig. 4 shows negative covariance. The colored-ball analogy the author invokes points the same way: counting colors from N draws gives multinomial counts, which are negatively correlated, not independent. So Eq. (44), the product-form PDF for b0, and the Binomial-sum distributions for b1, χ, and bsum rest on an assumption that the author's own data falsify. The later claim that cross-correlation corrections to means and variances are 'likely small' is hard to square with the fact that the negative covariance visibly changes σ_χ and σ_bsum (Eqs. 28–29).\n\nThe CLT section is heuristic. It invokes block-sum independence and boundedness but never verifies that the variance sums diverge or that the block sums are actually independent. The result is plausible, but it is a sketch, not a theorem, and it does not engage with the rigorous CLTs already in Owada–Thomas and Owada–Samorodnitsky (refs 56/57) beyond citing them.\n\nThe Binomial validation is partly circular in the way the reader says: N and p_j are calibrated from the ensemble mean and variance of each statistic, and the same numerical PDFs are then used for comparison. That leaves only the skewness/kurtosis as a true test, which is weaker than the paper implies. A two-parameter Binomial matching mean and variance can look reasonable even when the independence structure is wrong.\n\nThere are also small slips: the text says Binomial corresponds to drawing without replacement (it is the opposite; hypergeometric is without replacement), and there is a typo in the binomial coefficient in Eq. (47).\n\nWho is this for? Cosmologists using Betti numbers for error bars would benefit from a correct version of this, but they should not yet trust the Binomial-sum formulas. The paper deserves a serious referee because the problem is important, the numerical work is substantial, and the representation could be the starting point for a proper treatment. I would send it to review, but with the expectation that the distributional claims need to be re-derived with dependence included (e.g., multinomial) or substantially weakened. As it stands, reject.","headline":"A clean bookkeeping device plus a useful observation about bsum's stability, but the Binomial independence model contradicts the paper's own measured negative covariance and the CLT is heuristic; deserves referee time, but the distributional claims should not survive review as they stand.","tokens_in":18355,"tokens_out":5998,"would_cite":false,"duration_ms":68475,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60G60","60F05","62E20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Betti numbers of smooth random fields can be read as sums of independent coin-flip-like counting variables.","keywords":["Betti numbers","Euler characteristic","excursion sets","Binomial distribution","Gaussian random fields","topological statistics","central limit theorem","cosmology"],"falsifier":"Simulate many realizations of a zero-mean Gaussian field on a fixed grid, compute the excursion sets at a threshold, and count the basis coefficients $m_0,m_1,\\ldots$ across realizations. If the sample covariance of, say, $m_0$ and $m_1$ is systematically nonzero beyond Monte Carlo noise, the independence assumption fails and the predicted product-form PDFs do not hold. Alternatively, at a threshold where $N\\simeq 1$ (e.g., $\\nu>3$), the Binomial model predicts variance $Np(1-p)$; a measured variance that departs from this with the same mean signals a misspecified model.","tokens_in":17377,"feed_emoji":"🎲","tokens_out":5094,"duration_ms":51182,"temperature":0.7,"pith_summary":"This paper claims that the random topology of a smooth random field can be understood through a small set of counting variables. Any excursion set—the region where the field exceeds a threshold—is written as a union of basic building blocks (disks, disks with holes, and so on), and the numbers $m_j$ of each building block are modelled as independent Binomial variables. Betti numbers, the Euler characteristic, and the sum of Betti numbers are then literal sums of these $m_j$'s, which makes their probability distributions follow from simple combinatorial ones and become Gaussian when the field of view divided by the correlation length is large. If true, this gives cosmologists a principled reason for when Gaussian error bars on topological statistics are trustworthy and when they are not.","feed_headline":"Betti numbers of random fields follow from coin flips","feed_subtitle":"A new topological basis turns excursion-set counts into Binomial sums, explaining when Gaussian error bars are safe.","key_machinery":"The topological basis $B=\\{Q_0,Q_1,Q_2,\\ldots\\}$, where each $Q_j$ is a connected component with exactly $j$ holes (in three dimensions, additional indices track handles and cavities), together with a colored-balls-in-a-box model that realizes each coefficient $m_j$ as the count of draws of color-$j$ balls in $N$ trials, i.e., a Binomial variable. The identities $b_0=\\sum_j m_j$, $b_1=\\sum_j j m_j$, $\\chi=\\sum_j(1-j)m_j$, and $b_{\\rm sum}=\\sum_j(1+j)m_j$—obtainable from the generating function $h(\\alpha)=\\sum_j m_j e^{-j\\alpha}$—are what carry the argument, because they reduce the statistical question about topology to the classical central limit theorem for sums of independent discrete variables.","core_discovery":"The paper's central claim is that the coefficients $m_j$ of the topological basis are Binomial random variables. With $b_0 = \\sum_j m_j$, $b_1 = \\sum_j j m_j$, $\\chi = \\sum_j (1-j)m_j$, and $b_{\\rm sum} = \\sum_j (1+j)m_j$, the probability distribution of each topological statistic is determined by the product of Binomial distributions of the $m_j$'s. Assuming independence and a growing number of building blocks, a central limit argument shows that $b_0,b_1,\\chi,b_{\\rm sum}$ become Gaussian at each threshold as $q\\equiv (L/r_c)^d\\to\\infty$. The paper tests this model on Gaussian temperature maps and finds the Binomial and Gaussian descriptions fit the simulated PDFs well except at high thresholds where the effective sample size $N$ is of order one.","pith_inferences":["If the Binomial model holds, the full distribution of any topological statistic is determined by its mean and variance at each threshold; this two-parameter description could let observers convert a measured value into a statistical significance without running simulations.","The independence assumption could be relaxed to block independence; a direct measurement of covariances between $m_j$'s on synthetic fields would settle whether the paper's variance formulas need correction terms.","Because the basis coefficients count components by their number of holes, the same Binomial logic may constrain the fluctuations of persistence diagrams, whose birth-death counts share a similar combinatorial structure.","The paper claims the Binomial form is general, but its numerical test is limited to Gaussian fields; checking the model on lognormal or Rayleigh random fields is a direct next test of the claim."],"forward_implications":["Topological statistics are asymptotically Gaussian at each threshold provided $q\\to\\infty$, with the convergence rate depending on threshold through $j_{\\max}(\\nu)$.","At intermediate thresholds, the PDFs of $b_0,b_1,\\chi,b_{\\rm sum}$ are well approximated by Binomial or Gaussian models, while at $|\\nu|>3$, where $N$ is of order one, the Gaussian approximation breaks down.","The anti-correlation between $b_0$ and $b_1$ makes $b_{\\rm sum}$ have lower variance than either, which the paper notes could yield tighter cosmological parameter constraints.","The Binomial model explains the threshold dependence: at $\\nu>1$, $b_0\\sim\\chi\\sim b_{\\rm sum}\\sim m_0$ with $p\\sim p_0$, while at $\\nu<-1$, $b_1\\sim -\\chi\\sim b_{\\rm sum}\\sim j_{\\max}\\,m_{j_{\\max}}$.","The 3D generalization expresses $b_0,b_1,b_2,\\chi,b_{\\rm sum}$ as sums of coefficients $m_{ij_0j_1\\ldots}$, with the same Binomial modeling and the same asymptotic Gaussian behavior restricted to $q\\to\\infty$."],"supporting_citations":[{"why":"Supplies the definitions of random fields, excursion sets, and the analytic Gaussian expectation formulas used to calibrate the Binomial parameters.","marker":"[1]"},{"why":"Provides the closed-form analytic formula for $\\langle\\chi\\rangle$ in Gaussian fields that anchors the calibration of $p_0$ and $N$.","marker":"[47]"},{"why":"Establishes numerical Betti numbers for Gaussian random fields, which the paper uses as a baseline for its own numerical comparisons.","marker":"[10]"},{"why":"Supplies the broader numerical study of Betti numbers that motivates why the statistics are hard to treat analytically.","marker":"[11]"},{"why":"Formulates the Lindeberg central limit theorem for independent discrete variables, which the paper invokes for the Gaussian convergence of $b_0,b_1,\\chi,b_{\\rm sum}$.","marker":"[63]"},{"why":"Gives the classical CLT statement and boundedness condition used to justify asymptotic Gaussianity.","marker":"[64]"},{"why":"Provides Alexander duality, which underlies the symmetry $b_0(\\nu)=b_1(-\\nu)$ used in the threshold-regime analysis.","marker":"[67]"},{"why":"Supplies the stars-and-bars combinatorial counting used to compute the number of coefficient states for fixed $(b_0,b_1)$.","marker":"[69]"}],"fun_headline_variants":["Topology of random fields from Binomial sums","Betti numbers as coin tosses","Excursion-set topology: Binomial to Gaussian","Random field topology: Gaussian when many building blocks","Cosmic topology follows Binomial statistics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole distributional story rests on the assumption that the coefficients $m_j$ are statistically independent Binomial variables with parameters $(p_j,N)$ set by the field; the paper does not derive this from the field's probability structure.","fun_headline_variants_meta":{"raw":{"variants":["Topology of random fields from Binomial sums","Betti numbers as coin tosses","Excursion-set topology: Binomial to Gaussian","Random field topology: Gaussian when many building blocks","Cosmic topology follows Binomial statistics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000178,"raw_usage":{"total_tokens":1289,"prompt_tokens":928,"completion_tokens":361,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":544,"completion_tokens_details":{"reasoning_tokens":294}},"tokens_in":544,"tokens_out":361,"duration_ms":4150,"temperature":1.0,"reasoning_tokens":294,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:43:15.556699+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate many realizations of a zero-mean Gaussian field on a fixed grid, compute the excursion sets at a threshold, and count the basis coefficients $m_0,m_1,\\ldots$ across realizations. If the sample covariance of, say, $m_0$ and $m_1$ is systematically nonzero beyond Monte Carlo noise, the independence assumption fails and the predicted product-form PDFs do not hold. Alternatively, at a threshold where $N\\simeq 1$ (e.g., $\\nu>3$), the Binomial model predicts variance $Np(1-p)$; a measured variance that departs from this with the same mean signals a misspecified model.","supporting_citations":[{"cited_title":"An example is a power law form with spectral index α","cited_arxiv_id":null,"evidence_quote":"Supplies the definitions of random fields, excursion sets, and the analytic Gaussian expectation formulas used to calibrate the Binomial parameters."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes numerical Betti numbers for Gaussian random fields, which the paper uses as a baseline for its own numerical comparisons."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the broader numerical study of Betti numbers that motivates why the statistics are hard to treat analytically."},{"cited_title":"Owada and A","cited_arxiv_id":null,"evidence_quote":"Formulates the Lindeberg central limit theorem for independent discrete variables, which the paper invokes for the Gaussian convergence of $b_0,b_1,\\chi,b_{\\rm sum}$."},{"cited_title":"Owada and G","cited_arxiv_id":null,"evidence_quote":"Gives the classical CLT statement and boundedness condition used to justify asymptotic Gaussianity."},{"cited_title":"We are not considering scaling transformations of the manifold","cited_arxiv_id":null,"evidence_quote":"Provides Alexander duality, which underlies the symmetry $b_0(\\nu)=b_1(-\\nu)$ used in the threshold-regime analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the stars-and-bars combinatorial counting used to compute the number of coefficient states for fixed $(b_0,b_1)$."}],"review_version":1}