{"id":"820adc36-d511-4055-82eb-3f5970bd538f","arxiv_id":"2411.09621","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper reviews the bridge between Bernoulli percolation and Galton-Watson trees and introduces Divide-and-Color percolation as a representation of finite-allele neutral mutation models, with a phase-transition theorem.","lead":"This survey connects percolation theory with branching processes by showing that two kinds of neutral mutation models on Galton-Watson family trees correspond to two color-percolation models. It adds a threshold formula for when an infinite family of one type survives under Divide-and-Color percolation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.2 survives: the existence equivalence in Prop. 4.5 suffices for the threshold and the exploration coupling is valid; the real defect is the MIM law in Eqs. (7)/(14) omitting the k=0 term.","rationale":"I examined the central claim in Theorem 5.2. The proof's first sentence asserts that the DaC same-color component of the root is distributed as a Bernoulli(ep) percolation cluster on the BGW tree. This assertion is not proved in the paper, but it is true: conditioning on the tree, the exploration of the DaC component gives each child of an included vertex an independent inclusion probability ep = p + (1-p)a, and the recursive structure is exactly that of a Bernoulli(ep) cluster. Moreover, the theorem's threshold formula does not even require full distributional equality; Proposition 4.5's equivalence of the events of infinite size, combined with Theorem 5.1, is sufficient. Thus the reader's weakest assumption is a proof gap but not a correctness gap. The paper does contain a genuine error in the MIM model: Eq. (7) and Eq. (14) sum the number of clones k from 1 to v_i, omitting k = 0. As a result, the printed MIM law assigns probability zero to the event that a mother has no same-type offspring, contradicting both the verbal description and the DaC correspondence. This error is localized and easily fixed, so it does not change the verdict: CONDITIONAL remains appropriate. I recommend that the authors correct the summation limits and add a short exploration argument for the distributional identification in Theorem 5.2.","tokens_in":20177,"tokens_out":25606,"duration_ms":257076,"concrete_test":"Recompute Eq. (14) with u_i = 0: the current sum over k = 1 to u_i is empty and gives probability 0, whereas the intended mother-independent mechanism gives probability r^{|u|}(1/d)^{|u|} > 0 that all |u| children are mutants and all choose type i. Change the lower limit to k = 0 and verify that the resulting conditional law matches the one-generation DaC measure with p = 1 - r and a = (1/d, ..., 1/d); if the corrected expression agrees, the MIM/DaC correspondence is restored.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's weakest assumption is not actually load-bearing for Theorem 5.2. Proposition 4.5 already gives, for each fixed tree, the equivalence P(DaC root-component is infinite) > 0 iff P(Bernoulli(ep) root-cluster is infinite) > 0, with ep = 1 - (1-p)(1-a_empty). Since Theorem 5.1 says the latter occurs iff ep > 1/m on a BGW tree conditioned on non-extinction, solving p + (1-p)a_empty = 1/m gives exactly p_c = (1 - m a_empty)/(m(1 - a_empty)). Full distributional equality is not needed for the threshold formula, and in any case a standard exploration coupling shows the vertex set of the DaC same-color component is indeed distributed as the Bernoulli(ep) cluster, so the compressed proof is correct though terse. The concrete defect I find is in the MIM model definition: Eq. (7) and Eq. (14) sum k from 1 to v_i, omitting k = 0. The described mechanism allows a mother to have zero same-type children with positive probability, and the DaC model assigns positive probability to this event; as printed, the MIM law assigns probability 0 to v_i = 0, so the claimed coincidence with the DaC measure in Section 5.1.2 is false for the printed definition. This does not invalidate Theorem 5.2, but it makes a stated central correspondence incorrect until the lower summation limit is fixed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper is a survey of connections between percolation models and Bienaymé–Galton–Watson (BGW) branching structures. It reviews Bernoulli bond percolation on trees, the infinite-allele neutral mutation model, and Haggström's Divide-and-Color (DaC) percolation. Its main new contribution is Theorem 5.2, which claims that for a BGW tree with mean offspring number m>1 and conditioning on non-extinction, the critical DaC percolation parameter is ep_T^c = (1 - m a_∅)/(m(1 - a_∅)) almost surely. The paper also introduces two finite-allele mutation models, the mother-dependent model (MDM) and the mother-independent model (MIM), and claims a distributional correspondence between MIM and DaC percolation and between MDM and a restricted DaC model.","tokens_in":20483,"tokens_out":3462,"duration_ms":33899,"significance":"If Theorem 5.2 is correct, it genuinely extends Haggström's phase-transition result from deterministic trees to random BGW trees conditioned on non-extinction, and it gives a phase-transition threshold for a finite-allele neutral mutation model. The theorem is obtained by combining external known results—Lyons' p_c = 1/m for BGW trees and Haggström's exploration equivalence—rather than by a fully self-contained proof, but the claimed threshold is falsifiable and, as the stress-test discussion confirms, the proof can be made rigorous through Proposition 4.5 without full distributional equality. The expository portions of the paper are well organized and useful. No machine-checked proofs or code are provided, but the main result is elementary enough that this is not a barrier.","major_comments":[{"comment":"The summation in the MIM offspring distribution starts at k=1, where k denotes the number of clone children. The case k=0 must be included: a mother can have zero clone children, and both the mechanism described in §3.4.2 and the DaC model assign positive probability to that event. As printed, Eq. (7) gives probability 0 to every configuration with v_i=0, so the claimed coincidence with the DaC measure in §5.1.2 is false as written. The lower summation limit should be changed from 1 to 0.","section":"§3.4.2, Eq. (7); §5.1.2, Eq. (14)"},{"comment":"The proof states that the same-color connected component containing a given vertex 'is distributed as in Bernoulli percolation' with parameter ep, but it gives no proof of this distributional identification for the random-tree case. As written, Proposition 4.5 only establishes equivalence of positive probability of an infinite component, not full distributional equality. Since only the threshold is needed, the equivalence in Proposition 4.5 combined with Theorem 5.1 suffices to derive the stated formula; the proof should be expanded along those lines, explicitly invoking the exploration coupling or citing it precisely in the BGW setting.","section":"Theorem 5.2, proof in §5.2.2"}],"minor_comments":[{"comment":"The symbol a_∅ is used in the statement and proof but is not defined in the proposition; it should be defined as the probability that a percolation cluster receives the same color as the root (i.e., a_1 in the two-color case).","section":"§4.3.1, Proposition 4.5"},{"comment":"The interpretation of Theorem 5.2 for the MIM model should display the substitutions p = 1-r and a_∅ = 1/d before writing conditions in terms of 1-r, so that the comparison between the mutation parameter and ep_T^c is explicit.","section":"§5.2.2, Eqs. (21)–(22)"},{"comment":"References [25] and [26] contain the typo 'Phyisica' instead of 'Physica'; the spelling of Haggström should also be made consistent throughout.","section":"References"},{"comment":"The notation for the number of offspring of a vertex is ku(t) in §2.2 but k∅(T) in §3.1; using one convention throughout would improve readability.","section":"§2.2 and §3.1"}],"recommendation":"major_revision","confidential_remarks":"The indexing error in Eqs. (7) and (14) is a typo-level defect that nevertheless makes a stated central correspondence incorrect as printed. The main threshold theorem appears defensible and can be repaired by expanding the proof via Proposition 4.5 and Theorem 5.1. I see no concerns about novelty, citation practice, or fit with the journal's scope, assuming the corrections are made."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new content here is the correspondence between Divide-and-Color percolation and the two finite-allele mutation models (MIM and MDM), plus Theorem 5.2 giving the DaC critical value on supercritical Galton-Watson trees. The rest is expository. That is not a criticism: the survey parts are clean and the gentle treatment of Bertoin's and Haggstrom's results is well done. The paper earns its keep as a bridge between two literatures, and the new claims are real, not just repackaged old results.\n\nThe main theorem survives scrutiny. The worry that Theorem 5.2 needs full distributional equality between the DaC same-color component and Bernoulli(ep) percolation is not load-bearing: Proposition 4.5 already gives the existence equivalence, and with Lyons' threshold that is enough to derive ep_c = (1 - m a_empty)/(m(1 - a_empty)). The exploration coupling that justifies the distributional statement is standard, so the compressed proof is terse but sound.\n\nThe actual defect is the MIM definition. Equations (7) and (14) sum k from 1 to v_i, dropping the k=0 term. The mechanism explicitly allows a mother to have zero same-type children with positive probability, and the DaC model gives that event positive probability. As printed, the MIM law assigns zero probability to v_i = 0, so the claimed coincidence with the DaC measure in Section 5.1.2 is false. This is a one-line fix (lower the sum to k=0), but it is load-bearing for the stated correspondence, so it must be corrected before publication.\n\nMinor point: the proof of Theorem 5.2 cites \"Proposition 5.9 in [Lyons-Peres]\" without checking that the proposition's conditions hold on a BGW tree conditioned on non-extinction. This is probably harmless because the relevant percolation on the tree is a branching process, but the citation deserves a line of justification.\n\nThe citation pattern is honest. The new results build on Haggstrom, Lyons, and Bertoin in exactly the way the text says, and the self-citations to Blancas-Rivero are background only.\n\nWho is this for? People working at the interface of percolation and branching processes, especially those interested in population-genetic interpretations. They will find the survey useful and the new theorem a modest but genuine extension. It is not a breakthrough, but it is a competent piece of work that deserves referee time. My recommendation: send it to review, with the request that the authors fix the summation limits in the MIM law and expand the one-line justification in Theorem 5.2's proof.","headline":"A mostly survey paper with a genuinely new DaC/MIM correspondence and a threshold theorem that is correct in substance but printed with an indexing error in the MIM law.","tokens_in":21032,"tokens_out":1157,"would_cite":false,"duration_ms":12744,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J80","92D25","60K35","82B43"],"pacs":[],"model":"deepseek-v4-flash","headline":"The critical Divide-and-Color percolation threshold on a supercritical Galton-Watson tree, conditioned on non-extinction, is $(1 - m a_\\emptyset)/(m(1 - a_\\emptyset))$, which locates the appearance of infinite same-type families in the…","keywords":["Bienaymé-Galton-Watson tree","Bernoulli bond percolation","Divide-and-Color percolation","neutral mutations","infinite alleles","finite alleles","phase transition","mother-independent mutation model"],"falsifier":"Take a BGW tree with offspring distribution $P(\\xi=0)=0.1$ and $P(\\xi=3)=0.9$, so $m=2.7$, set $a_\\emptyset=0.3$, and compute the claimed threshold $p=(1-2.7\\cdot 0.3)/(2.7\\cdot 0.7)\\approx 0.1005$. Use dynamic programming over the first $n$ generations to compute the probability that the root's same-color DaC component reaches depth $n$; the claimed distributional equivalence predicts survival probability $0$ at this $p$ and positive survival only for $p$ above it. If the recursion gives positive survival at the claimed threshold, the identification of the color component with a Bernoulli($\\tilde p$) percolation cluster on a random BGW tree fails.","tokens_in":19987,"feed_emoji":"🌳","tokens_out":20088,"duration_ms":162976,"temperature":0.7,"pith_summary":"This survey establishes a dictionary between percolation on Bienaymé-Galton-Watson (BGW) trees and neutral mutation models in population genetics. It shows that Bernoulli bond percolation with open-edge probability $1-r$ is the same construction as a BGW process with infinite neutral alleles, and that Divide-and-Color percolation describes a finite-allele, mother-independent mutation model. The novel result is a phase-transition formula: on a BGW tree with mean offspring $m>1$, conditioned on non-extinction, the critical Divide-and-Color parameter is $(1-m a_\\emptyset)/(m(1-a_\\emptyset))$, where $a_\\emptyset$ is the probability that a percolation cluster receives the root's color. A sympathetic reader would care because this turns a population-genetics question, when does an infinite same-type family appear, into a percolation threshold calculation.","feed_headline":"Same-type families turn infinite past a simple threshold","feed_subtitle":"On supercritical Galton-Watson trees the threshold is (1 − m a)/(m(1 − a)), locating finite-allele mutation transitions.","key_machinery":"The key machinery is the exploration argument for Divide-and-Color percolation on trees. It observes that, for a fixed percolation configuration, a child of a vertex either belongs to the same percolation cluster as its parent, which happens with probability $p$, or belongs to a different cluster independently assigned the parent's color, which happens with probability $(1-p)a_\\emptyset$; hence the child inherits the parent's color with probability $\\tilde p = p+(1-p)a_\\emptyset = 1-(1-p)(1-a_\\emptyset)$. This reduces the same-color connected component of the root to a Bernoulli bond-percolation cluster at parameter $\\tilde p$, and the known critical value $1/m$ for BGW trees conditioned on non-extinction then yields the formula by solving $\\tilde p = 1/m$ for $p$.","core_discovery":"The paper's central discovery is that two apparently different probability models are the same object viewed from two directions, plus a new threshold for one of them. On a BGW tree, Bernoulli bond percolation with parameter $1-r$ reproduces the allelic partition of a BGW process with infinite neutral alleles: open edges are clone edges and closed edges are mutations. For finitely many alleles, the paper proposes that Divide-and-Color percolation, bond percolation followed by independent random coloring of clusters, matches the mother-independent mutation (MIM) model, while the restricted Divide-and-Color model, in which a daughter cluster cannot inherit its mother's color, matches the mother-dependent model. Theorem 5.2 asserts that on a supercritical BGW tree conditioned on non-extinction, the root's same-color component is distributed as the cluster of the root in Bernoulli bond percolation at effective parameter $\\tilde p = 1-(1-p)(1-a_\\emptyset)$; consequently the critical value of $p$ is $(1-m a_\\emptyset)/(m(1-a_\\emptyset))$ almost surely, and the probabilities that an infinite same-type subtree exists and that the root's type-subtree is infinite both switch at this threshold.","pith_inferences":["The distributional identification in Theorem 5.2 is carried from deterministic trees to random BGW trees by analogy; if it holds only for the event of an infinite component rather than for the full law of the root's color cluster, the threshold derivation would need an extra step.","The formula implies that with $d$ uniform alleles the critical mutation rate is $r_c = d(m-1)/(m(d-1))$, which decreases as $d$ grows, so larger allele repertoires make an infinite same-type family harder to maintain and require a smaller mutation rate.","Because the threshold is a function of $m$ alone, the same formula should describe every BGW tree with the same mean offspring, giving a testable universal prediction for finite-allele mutation models on random genealogies."],"forward_implications":["For the mother-independent finite-allele model with $d$ equally likely alleles, so $a_\\emptyset=1/d$, an infinite same-type family appears if and only if $1-r$ exceeds $(1-m/d)/(m(1-1/d))$.","For the mother-dependent mutation model, the restricted Divide-and-Color construction gives the same $1/m$ critical condition, so its infinite-type phase transition is governed by the same Bernoulli-percolation threshold.","The infinite-allele neutral-mutation model and Bernoulli percolation on a BGW tree are the same construction, so percolation theorems on BGW trees translate directly into statements about infinite allelic subtrees.","Conditioned on non-extinction, the threshold depends only on the mean $m$ and the color probability $a_\\emptyset$, not on finer details of the offspring distribution."],"supporting_citations":[{"why":"Introduces the Divide-and-Color model and supplies the deterministic-tree equivalence between an infinite same-color component and Bernoulli percolation at the effective parameter.","marker":"[31]"},{"why":"Establishes that the critical Bernoulli percolation parameter on a BGW tree conditioned on non-extinction is $1/m$.","marker":"[48]"},{"why":"Provides the percolation-on-trees background and the result used to conclude that the color cluster is finite when the effective parameter times $m$ is at most $1$.","marker":"[49]"},{"why":"Introduces the BGW process with infinite neutral mutations and its allelic partition, the model that Bernoulli percolation on a BGW tree is shown to reproduce.","marker":"[7]"},{"why":"Develops the rare-mutation regime for the same neutral-mutation process, supplying the population-genetics setting the survey's bridge extends.","marker":"[8]"},{"why":"Gives the extinction and phase-transition criteria for BGW and multi-type BGW processes used throughout the survey.","marker":"[5]"}],"fun_headline_variants":["Percolation and genetics: a bridge on Galton-Watson trees","Divide-and-color percolation mirrors multi-type trees","A single threshold for infinite same-type families","Two models, one threshold: percolation meets genetics","New link between percolation and mutation on trees"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that on a random Galton-Watson tree the root's same-color Divide-and-Color component has exactly the distribution of a Bernoulli bond-percolation cluster at effective parameter $\\tilde p = 1-(1-p)(1-a_\\emptyset)$, with the critical value $1/m$ applying after conditioning on non-extinction; the paper imports this identification from the deterministic-tree case rather than proving it for random trees.","fun_headline_variants_meta":{"raw":{"variants":["Percolation and genetics: a bridge on Galton-Watson trees","Divide-and-color percolation mirrors multi-type trees","A single threshold for infinite same-type families","Two models, one threshold: percolation meets genetics","New link between percolation and mutation on trees"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001358,"raw_usage":{"total_tokens":5479,"prompt_tokens":885,"completion_tokens":4594,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":4514}},"tokens_in":501,"tokens_out":4594,"duration_ms":36143,"temperature":1.0,"reasoning_tokens":4514,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:27:41.738949+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a BGW tree with offspring distribution $P(\\xi=0)=0.1$ and $P(\\xi=3)=0.9$, so $m=2.7$, set $a_\\emptyset=0.3$, and compute the claimed threshold $p=(1-2.7\\cdot 0.3)/(2.7\\cdot 0.7)\\approx 0.1005$. Use dynamic programming over the first $n$ generations to compute the probability that the root's same-color DaC component reaches depth $n$; the claimed distributional equivalence predicts survival probability $0$ at this $p$ and positive survival only for $p$ above it. If the recursion gives positive survival at the claimed threshold, the identification of the color component with a Bernoulli($\\tilde p$) percolation cluster on a random BGW tree fails.","supporting_citations":[{"cited_title":"Haggstr¨ om","cited_arxiv_id":null,"evidence_quote":"Introduces the Divide-and-Color model and supplies the deterministic-tree equivalence between an infinite same-color component and Bernoulli percolation at the effective parameter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that the critical Bernoulli percolation parameter on a BGW tree conditioned on non-extinction is $1/m$."},{"cited_title":"Lyons and Y","cited_arxiv_id":null,"evidence_quote":"Provides the percolation-on-trees background and the result used to conclude that the color cluster is finite when the effective parameter times $m$ is at most $1$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the BGW process with infinite neutral mutations and its allelic partition, the model that Bernoulli percolation on a BGW tree is shown to reproduce."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Develops the rare-mutation regime for the same neutral-mutation process, supplying the population-genetics setting the survey's bridge extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the extinction and phase-transition criteria for BGW and multi-type BGW processes used throughout the survey."}],"review_version":1}