{"id":"98de06a4-3806-4fea-a1b6-936922cab131","arxiv_id":"2412.16035","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A general invariance principle shows critical spatial branching processes converge to the Brownian CRT, proved with a new many-to-few moment formula.","lead":"Branching processes where particles have types that influence reproduction produce family trees that converge to the Brownian Continuum Random Tree after rescaling, under a moment condition. The paper introduces a new many-to-few formula that computes tree moments via a Markov chain on a uniform tree.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.2's induction uses m_2 as a continuous test function, but Assumption 1 only gives continuity of E|Ξ_x|^2; this gap affects the central moment asymptotics.","rationale":"Good-faith reading: the paper gives a coherent moment method and the many-to-few formula is a genuine contribution. The central claim is Theorem 1.1(i), and its proof hinges on Lemma 5.2. I examined Assumption 1(ii) but it is an explicit, natural condition; the more fragile point is the hidden continuity of m_2 in the induction. This is not an external-consensus issue but an internal proof gap: the test functions available from the induction are continuous bounded, while m_2 is only bounded. It can likely be repaired by strengthening Assumption 1(iii) to continuity of m_2, which holds in the finite-type and diffusion examples. Hence the appropriate verdict is CONDITIONAL rather than ACCEPT or REJECT. The reader's weakest_assumption (Assumption 1(ii)) is different; while also load-bearing, it is stated as an assumption rather than an unproven step, so I disagree with the reader's choice of the single weakest point.","tokens_in":32575,"tokens_out":35464,"duration_ms":338131,"concrete_test":"Independently re-derive the k=2 case of Lemma 5.2 under only Assumption 1, keeping the branch-point function m_2 arbitrary bounded. If the derivation requires continuity of m_2 (or of the map x↦E_x[Σ_{i≠j}h(ξ_i)h(ξ_j)]), then either add that continuity to Assumption 1(iii) and re-verify Proposition 5.1, or produce a counterexample: a branching process satisfying Assumptions 1(i)-(ii) and 1(iii) as stated on a non-discrete type space whose h-weighted factorial second moment m_2 is discontinuous at a point charged by π. In the counterexample case, the moment limit formula will fail for some continuous bounded test functions, so Theorem 1.1(i) is false as stated.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In the proof of Theorem 1.1(i), Proposition 5.1 reduces the moment limits to Lemma 5.2. In the induction step of Lemma 5.2 (from k to k+1 leaves), after conditioning on the reduced tree eτ_n = span{v_1,...,v_k, w}, w=v_k∧v_{k+1}, the conditional expectation of the two external branches converges uniformly to (1/2)m_2(X_w)⟨π,f_k⟩⟨π,f_{k+1}⟩, where m_2(y)=E_y[Σ_{i≠j}h(ξ_i)h(ξ_j)]. The proof then applies the induction hypothesis to eτ_n with a product-type leaf functional containing m_2(X_w). But the induction hypothesis (and the initial reduction to product functions) only covers continuous bounded f_i. Assumption 1(iii) guarantees that x↦E|Ξ_x|^2 is continuous and bounded, hence m_2 is bounded, but not that m_2 is continuous: the h-weighted factorial second moment depends on the joint law of the offspring point process, not only on its total number. Without continuity of m_2, or an additional argument proving the k-leaf biased leaf marginals converge weakly enough to evaluate m_2, the step is not justified. Since Lemma 5.2 feeds directly into Proposition 5.1 and hence the identification of the limit as the Brownian CRT moment, this is a load-bearing gap in the central claim as written.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves an invariance principle for the genealogical tree of a critical spatial branching process. Under Assumption 1 (existence of a harmonic function and stationary measure for the mean semigroup, uniform ergodicity of the one-particle semigroup, and uniform moment bounds), Theorem 1.1 identifies the vague Gromov-vague limit of the rescaled marked tree n L_x(\\bar T_n) as h(x) times the law of a free Brownian CRT with variance \\Sigma^2 and independent marks distributed according to \\pi. It also gives the corresponding limit for the generation-n ultrametric tree conditioned to have macroscopic size. Under the additional Kolmogorov-type estimate in Assumption 2, Theorem 1.2 reinforces the convergence to the marked Gromov-Hausdorff-Prohorov topology. The proof introduces a new many-to-few formula expressing k-th moments in terms of a Markov chain indexed by a uniform planar tree, then derives moment asymptotics via an induction on the number of leaves, and finally applies a method of moments for random metric measure spaces.","tokens_in":32838,"tokens_out":19143,"duration_ms":178839,"significance":"This is a substantial contribution to the scaling-limit theory of spatial branching processes. The many-to-few formula of Theorem 3.1 is a new and potentially reusable tool, and the moment approach provides a unified route to results that previously required model-specific encodings. The assumptions are explicit and cover multitype branching processes, branching diffusions in bounded domains, and other examples. The paper is careful with the infinite-measure formalism for Brownian CRT limits, and the statement of the limiting object in terms of \\Sigma^2 and \\pi is parameter-free. If the gap discussed below is repaired, the result would be an important step toward a general theory of genealogical convergence for spatially dependent branching mechanisms.","major_comments":[{"comment":"The induction step of Lemma 5.2 has a load-bearing gap. After conditioning on the reduced tree eτ_n, the proof applies the k-leaf induction hypothesis to a functional of the leaves of eτ_n that includes B_n(X_w), where B_n(y) = E_y[Σ_{i≠j} h(ξ_i) E_{ξ_i}[f_k(ζ)/h(ζ)] h(ξ_j) E_{ξ_j}[f_{k+1}(ζ)/h(ζ)]]. The proof shows that B_n(y) converges uniformly to (1/2)⟨π,f_k⟩⟨π,f_{k+1}⟩ m_2(y), and then invokes the induction hypothesis. However, the induction hypothesis is stated only for continuous bounded leaf functions, and Assumption 1(iii) guarantees continuity of x ↦ E|Ξ_x|^2, not of m_2(x) = E_x[Σ_{i≠j} h(ξ_i)h(ξ_j)] or of B_n. Without an additional approximation argument, the induction step is not justified. Since Lemma 5.2 feeds directly into Proposition 5.1 and hence into the identification of the limit as the Brownian CRT moment, this affects the central claim of Theorem 1.1. The proof can be repaired by strengthening Assumption 1(iii) to require continuity of m_2 (or of the relevant joint offspring moments), or by supplying an approximation argument that avoids evaluating m_2 as a continuous test function.","section":"Section 5.1, Lemma 5.2"}],"minor_comments":[{"comment":"The sentence 'Since the trees are converging deterministically, it is sufficient to prove the result for functionals that only depend on the types of the leaves' is terse; the passage from product functions to arbitrary continuous bounded F requires a monotone-class or tightness argument because the mark space E is not compact. The step is likely valid, but it deserves a short justification.","section":"Section 5.1, Lemma 5.2"},{"comment":"The induction proving the bound (36) is only sketched ('It will follow by an induction on k'). Since this bound is used to control the error terms B_{x_n,n} and eB_{x_n,n}, the induction should be written out or at least the base case and inductive step should be indicated explicitly.","section":"Section 5.2, Proposition 5.3"},{"comment":"The inequality n P_x(Σ_{m≥nR} Z_m ≥ ε n^2) ≤ n P_x(Z_{nR} > 0) is correct because the event on the left implies Z_{nR} > 0, but this implication should be stated for clarity, especially since the sum starts at m = nR rather than m > nR.","section":"Section 5.4, Corollary 5.4"},{"comment":"The notation T(2δn) and T_v^{2δn} is introduced quickly; the superscript convention (height truncation versus subtree truncation) could be made explicit to avoid confusion.","section":"Section 5.3, Lemma 5.7"}],"recommendation":"major_revision","confidential_remarks":"The main gap in Lemma 5.2 is local and likely fixable by adding continuity of the factorial second moment m_2 to Assumption 1(iii) or by a separate approximation lemma. I do not see evidence of circularity: the reliance on the author's preprint [30] for the method-of-moments framework is a technical dependency, not a circular use of the target result. The paper fits the scope of the journal and, once the gap is addressed, would be a strong contribution. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a substantial paper. It proves a Brownian CRT invariance principle for a broad class of critical spatial branching processes with type-dependent offspring, using a new many-to-few formula. The result genuinely unifies earlier special cases such as multitype Galton-Watson trees and branching diffusions in bounded domains. The many-to-few formula at all times and for arbitrary tree shapes is the real contribution, and the moment computation leading to the CRT moments is elegant. I think the paper deserves serious refereeing, but one technical point in the proof of Lemma 5.2 is not fully justified as written.\n\nWhat is new and good: the many-to-few formula (Theorem 3.1) is proven cleanly by induction and gives a tree-indexed Markov chain representation of moments. The method of moments is adapted carefully from earlier work. The assumptions (Assumption 1) are explicit, and the examples are sensible. The paper is well organized, and the bibliography is honest.\n\nSoft spot: in the induction step of Lemma 5.2, after conditioning on the reduced tree, the induction hypothesis is applied to a leaf functional containing m_2(X_w), where m_2(y) = E_y[sum_{i≠j} h(ξ_i)h(ξ_j)]. Assumption 1 guarantees m_2 is bounded but not continuous, and the induction hypothesis only covers continuous bounded functions. Without continuity of m_2, or a separate argument showing the biased leaf marginals can evaluate m_2, the step does not go through. This is load-bearing: Lemma 5.2 feeds directly into Proposition 5.1 and hence the identification of the limit as the Brownian CRT. The gap seems fixable: either add continuity of the second factorial h-moment to Assumption 1, or approximate m_2 by continuous functions and justify the limit with the uniform bounds in (32). But as written, it is a missing piece.\n\nMinor: the paper relies on the author's own preprint [30] for the method of moments. That is acceptable because the framework is cited, not the target result. The proof of the uniform bound (32) is a bit compressed in places, but that is a presentation issue, not a substantive gap.\n\nFor whom: probabilists working on branching processes, random trees, and scaling limits. It gives a broadly applicable toolkit. My recommendation: engage with it, but send to a serious referee and ask the author to address the Lemma 5.2 gap, either by strengthening the assumption or by supplying the missing approximation argument.","headline":"A genuinely new many-to-few formula and a broad Brownian CRT invariance principle, but the proof of Lemma 5.2 has a load-bearing gap that needs a fix or a stronger assumption.","tokens_in":33386,"tokens_out":4009,"would_cite":true,"duration_ms":34635,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J80","60F17","60B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the rescaled genealogy of a wide class of critical spatial branching processes converges to the Brownian Continuum Random Tree, with the limit determined only by a spatial variance parameter and the equilibrium…","keywords":["spatial branching processes","Brownian Continuum Random Tree","many-to-few formula","method of moments","Gromov-vague topology","coalescent point process","spine decomposition","critical branching processes"],"falsifier":"Take a critical branching process satisfying Assumptions 1(i) and (iii) but not (ii), for example a branching diffusion on an unbounded domain where the mean semigroup converges only pointwise rather than uniformly. Compute the fourth moment of the rescaled tree and compare it with the Brownian CRT moment formula of Proposition 5.1: a mismatch would show that the claimed universality fails exactly where uniform ergodicity fails.","tokens_in":32352,"feed_emoji":"🌳","tokens_out":9776,"duration_ms":81743,"temperature":0.7,"pith_summary":"This paper proves a universality theorem for the genealogical tree of a critical spatial branching process. Starting from any type $x$, the rescaled marked tree $\\bar T_n$, with edge lengths divided by the number of generations and particle masses by $n^2$, converges in the marked Gromov-vague topology to $h(x)$ times the law of a free Brownian Continuum Random Tree whose leaves carry independent marks with equilibrium distribution $\\pi$; the only quantity of the offspring law that survives is a spatial variance $\\Sigma^2$ (Theorem 1.1(i)). The same moment computation yields a second limit: the reduced tree spanned by the $n$-th generation converges to the Brownian coalescent point process with the same parameters, and with a supplementary survival-probability assumption both limits strengthen to the Gromov-Hausdorff-Prohorov topology. The interest is that the tree shape and particle locations may be strongly dependent, so classical path-encoding techniques fail; the paper's moment approach applies to a general class of such processes. A new many-to-few formula expresses $k$-th moments of the tree through a Markov chain indexed by a uniform tree with $k$ leaves, reducing convergence of the genealogy to computing moments.","feed_headline":"Critical spatial branching trees converge to the Brownian CRT","feed_subtitle":"A many-to-few formula shows only the spatial variance and the equilibrium type law survive in the limit.","key_machinery":"The engine is the many-to-few formula (Theorem 3.1): for a weighting function $\\psi$, the $k$-th moment measure of the branching tree equals $\\psi(x)\\sum_{\\tau\\in T_k} Q^\\psi_{x,\\tau}[\\Delta^\\psi_k F]$, a sum over planar trees with $k$ leaves of an expectation under a Markov chain indexed by that tree, with an explicit bias factor $\\Delta^\\psi_k$. With $\\psi=h$, the harmonic function of Assumption 1, the chain becomes the spinal Markov chain, the $h$-transform of the mean semigroup, and the bias simplifies. Proposition 5.1 then shows that, under Assumption 1, the rescaled $k$-th moments converge to $h(x)(\\Sigma^2/2)^{k-1}\\int E[F(\\theta,(X_i)_{i\\le k})]\\,d\\Lambda_k(\\theta)$, exactly the Brownian CRT moment formula of Proposition 4.2, with the $X_i$ i.i.d. under $\\pi$; the method of moments (Proposition 4.1) converts this into vague Gromov-vague convergence.","core_discovery":"The central claim, stated as Theorem 1.1(i), is that under Assumption 1 the law $n L_x(\\bar T_n)$ converges vaguely in the marked Gromov-vague topology to $h(x)L(T_{b,\\pi})$, where $T_{b,\\pi}$ is the free Brownian CRT with variance $\\Sigma^2$ and independent marks $\\pi$. Equivalently, the rescaled genealogy of a critical spatial branching process is universal: the Brownian CRT emerges regardless of the detailed offspring law, provided the mean semigroup mixes uniformly and the offspring have finite moments. The same method gives Theorem 1.1(ii), convergence of the rescaled generation-$n$ reduced tree to the Brownian coalescent point process, and Theorem 1.2, which upgrades both convergences to the Gromov-Hausdorff-Prohorov topology when the Kolmogorov survival estimate $nP_x(Z_n>0)\\to 2h(x)/\\Sigma^2$ holds. The paper also deduces conditioned limits, such as survival for a long time or a forest started from many ancestors, as corollaries.","pith_inferences":["A direct extension of the argument would replace the Brownian CRT moment formula with the moments of an $\\alpha$-stable CRT; the many-to-few machinery could then prove $\\alpha$-stable invariance principles for spatial branching processes, a case the paper only touches through examples and conjecture.","Because Assumption 1(ii) is a property of the one-particle mean semigroup rather than of tree shapes, the method should apply to processes whose genealogies are far from Galton-Watson trees as long as this uniform ergodicity holds; the paper proves this for critical processes, but the moment computation itself is not tied to a Galton-Watson structure.","The many-to-few formula gives an explicit biased representation of the subtree spanned by $k$ uniformly sampled particles, which could be used directly for coalescent inference or for simulating genealogies without constructing the whole population tree."],"forward_implications":["Under Assumption 1, the full rescaled genealogy of a critical spatial branching process is universal: no matter the offspring law, the limit is the free Brownian CRT with variance $\\Sigma^2$ and equilibrium marks $\\pi$.","The reduced tree spanned by a single generation converges to the Brownian coalescent point process with the same parameters, giving a scaling limit for the ultrametric genealogy of the $n$-th generation.","Conditioning on survival to a long time $tn$, or starting from $n$ ancestors, yields the conditioned Brownian CRT and a Poisson forest of Brownian CRTs (Corollary 1.3), so the moment method covers conditioned and multi-root limits as corollaries.","Kolmogorov's survival estimate $nP_x(Z_n>0)\\to 2h(x)/\\Sigma^2$ (when Assumption 2 holds) upgrades Gromov-vague convergence to Gromov-Hausdorff-Prohorov convergence, making diameter and height functionals continuous in the limit.","Convergence holds uniformly in the initial type $x$, so initial conditions can vary with $n$ as long as the empirical measure of starting types converges."],"supporting_citations":[{"why":"Supplies the moment formula for the Brownian CRT that the limit moments must match (Proposition 4.2 is a reformulation of its Theorem 3).","marker":"[55]"},{"why":"Gives the spinal-decomposition moment method for a single generation and the Brownian CPP moments that this paper extends to the full tree.","marker":"[33]"},{"why":"Introduces the many-to-few technique on which the new all-times many-to-few formula (Theorem 3.1) builds.","marker":"[41]"},{"why":"Provides the infinite-measure method of moments and vague convergence used to turn moment asymptotics into Gromov-vague convergence.","marker":"[30]"},{"why":"A source of the criticality and ergodicity assumptions, and of the Yaglom-limit context under which Assumptions 1 and 2 are stated.","marker":"[38]"},{"why":"Supplies the asymptotic-moment framework for spatial branching processes from which the assumptions on the mean semigroup are borrowed.","marker":"[35]"},{"why":"Supplies the tightness criterion that upgrades Gromov-vague convergence to Gromov-Hausdorff-Prohorov convergence under Assumption 2.","marker":"[7]"},{"why":"Constructs the Brownian coalescent point process that appears as the generation-$n$ reduced-tree limit.","marker":"[67]"},{"why":"Defines the marked metric measure space framework and topology in which the convergence statements live.","marker":"[25]"}],"fun_headline_variants":["Spatial branching trees converge to Brownian CRT via moments","Critical spatial branching processes: Brownian CRT limit","Moment method yields Brownian CRT for spatial branching","Spatial branching genealogy converges to Brownian CRT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is uniform ergodicity of the mean semigroup (Assumption 1(ii)): no matter where the process starts, the expected number of descendants of a given type approaches $h(x)\\langle \\pi,f\\rangle$ at a rate independent of the starting type, and without that uniformity the spine chain need not converge to the equilibrium $h\\pi$.","fun_headline_variants_meta":{"raw":{"variants":["Spatial branching trees converge to Brownian CRT via moments","Critical spatial branching processes: Brownian CRT limit","Moment method yields Brownian CRT for spatial branching","Spatial branching genealogy converges to Brownian CRT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000181,"raw_usage":{"total_tokens":1283,"prompt_tokens":894,"completion_tokens":389,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":510,"completion_tokens_details":{"reasoning_tokens":325}},"tokens_in":510,"tokens_out":389,"duration_ms":3647,"temperature":1.0,"reasoning_tokens":325,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:51:10.765505+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a critical branching process satisfying Assumptions 1(i) and (iii) but not (ii), for example a branching diffusion on an unbounded domain where the mean semigroup converges only pointwise rather than uniformly. Compute the fourth moment of the rescaled tree and compare it with the Brownian CRT moment formula of Proposition 5.1: a mismatch would show that the claimed universality fails exactly where uniform ergodicity fails.","supporting_citations":[{"cited_title":"Asymptotic genealogy of a critical branching process","cited_arxiv_id":null,"evidence_quote":"Constructs the Brownian coalescent point process that appears as the generation-$n$ reduced-tree limit."},{"cited_title":"The uniform random tree in a brownian excursion","cited_arxiv_id":null,"evidence_quote":"Supplies the moment formula for the Brownian CRT that the limit moments must match (Proposition 4.2 is a reformulation of its Theorem 3)."},{"cited_title":"Convergence of genealogies through spinal decomposition with an application to population genetics","cited_arxiv_id":null,"evidence_quote":"Gives the spinal-decomposition moment method for a single generation and the Brownian CPP moments that this paper extends to the full tree."},{"cited_title":"The many-to-few lemma and multiple spines","cited_arxiv_id":null,"evidence_quote":"Introduces the many-to-few technique on which the new all-times many-to-few formula (Theorem 3.1) builds."},{"cited_title":"Yaglom limit for critical nonlocal branching markov processes","cited_arxiv_id":null,"evidence_quote":"A source of the criticality and ergodicity assumptions, and of the Yaglom-limit context under which Assumptions 1 and 2 are stated."},{"cited_title":"Asymptotic moments of spatial branching processes","cited_arxiv_id":null,"evidence_quote":"Supplies the asymptotic-moment framework for spatial branching processes from which the assumptions on the mean semigroup are borrowed."},{"cited_title":"The gap between Gromov-vague and Gromov–Hausdorff-vague topology","cited_arxiv_id":null,"evidence_quote":"Supplies the tightness criterion that upgrades Gromov-vague convergence to Gromov-Hausdorff-Prohorov convergence under Assumption 2."},{"cited_title":"Marked metric mea- sure spaces","cited_arxiv_id":null,"evidence_quote":"Defines the marked metric measure space framework and topology in which the convergence statements live."}],"review_version":1}