{"id":"724a3f65-ddc3-4748-babc-511aa278d002","arxiv_id":"2412.05709","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For the critical GFF level set and loop soup on Z^d with d != 6, four IIC definitions coincide, and the conditioned cluster volume in B(M) is of order M^{min(d/2+1, 4)}.","lead":"This paper proves that four natural ways of conditioning the critical Gaussian free field or loop soup on the metric graph to form a large cluster all converge to the same incipient infinite cluster, in every dimension except 6. It also proves the volume of the large cluster inside a box of side M typically grows like M to the power min(d/2+1, 4), matching the fractal dimension conjectured by Werner.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"External quasi-multiplicativity (1.18) from companion paper [5] is the load-bearing input: it underpins the Hopf contraction in Prop 3.2 and the volume lower bound in Lemma 5.1; if it fails, Theorems 1.1 and 1.2 fall.","rationale":"I read the paper in good faith. The internal argument is coherent: the decomposition into annuli, the control of large loops via Lemma 3.5, the approximation Lemma 3.6, and the Hopf contraction are executed carefully. I found no internal inconsistency and no fabricated entities. The one point on which the entire edifice turns is the quasi-multiplicativity (1.18), imported verbatim from the companion preprint [5]. The reader's weakest_assumption names exactly this, and I agree. My concrete check is to derive (3.95) for the Hopf kernel from (1.18) with explicit constants; this is the step that would settle the matter. Since the companion is not verified here and the main theorems (Theorem 1.1 and the volume lower bound in Theorem 1.2) collapse if (1.18) fails, I recommend a conditional acceptance rather than an unconditional ACCEPT. This is not an objection to the paper's internal logic; it is a request that the external black box be opened. The paper itself flags d=6 as excluded precisely because of the missing quasi-multiplicativity, which reinforces that this is the central dependency.","tokens_in":52352,"tokens_out":18530,"duration_ms":161001,"concrete_test":"Prove or disprove the following condition: for every k≥0 and every admissible D_k⊂eB(n_{6k+4}), D_{k+1}⊂[eB(n_{6k+6})]^c, the kernel K_k(D_k,D_{k+1}) = P^{(qD_k∪qD_{k+1})}(D_k ← → D_{k+1}) satisfies (3.95) with a constant κ depending only on d. This should be done by re-deriving (1.18) from [5, Theorem 1.1] and verifying the inclusions required for (1.18) at the intermediate scale N=n_{6k+5}; if the ratio in (3.95) is bounded uniformly in k, the Hopf contraction step is sound, otherwise Proposition 3.2 fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Section 3.3 (Case 2) proves the uniform convergence in Proposition 3.2 by applying Hopf's contraction theorem (Lemma 3.7) to the kernels K_k(D_k,D_{k+1}) = P^{(qD_k∪qD_{k+1})}(D_k ← → D_{k+1}) in the integral I below (3.75). The condition (3.95) for these kernels is asserted to follow from the quasi-multiplicativity bound (1.18) imported from [5, Theorem 1.1]. This is the single most load-bearing input: (1.18) is not proved here, it carries a nontrivial correction factor N^{0⊡(6-d)} for d>6, and the kernel's absorbing boundary qD_k∪qD_{k+1} must satisfy the hypotheses of (1.18) uniformly in k. If the factorization in (1.18) failed, the contraction ratio in (3.100) would not be uniformly bounded by (κ−1)/(κ+1), and the uniform convergence of P(A_h | H^⋄_h) would collapse, taking Theorem 1.1 with it. The same bound is used in (5.48) via Lemma 5.1 to prove the volume lower bound of Theorem 1.2. Thus both main theorems rest on an external factorization that is structurally different from the IIC existence claim and is unverified in this paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes, for the critical Gaussian free field level-set and the critical loop soup on the metric graph \\tilde Z^d with d ≥ 3 and d ≠ 6, the existence and equivalence of four incipient infinite cluster (IIC) constructions: conditioning on connection to ∂B(N), on supercritical connection to infinity as the parameter tends to criticality, on connection to a distant lattice point, and on the capacity of the critical cluster exceeding T. The common IIC is shown to be almost surely infinite and one-ended. The paper further proves that under the conditioning {0 ↔ ∂B(N)} with N ≫ M, the volume of the critical cluster inside B(M) is typically of order M^{(d/2+1) ∧ 4}, and that the tail probability for large volume is bounded below by a constant depending only on d and the multiplier; these results are transferred to the IIC measure by taking N → ∞. The proofs adapt the Basu–Sapozhnikov framework to the loop-soup representation, with the key quasi-multiplicativity estimate imported from the authors' companion paper [5].","tokens_in":52604,"tokens_out":10885,"duration_ms":105205,"significance":"If correct, the paper is a substantial advance: it unifies four natural IIC notions in a strongly correlated, non-Bernoulli percolation model, gives the first rigorous volume-growth order for the critical cluster under a large-diameter conditioning, and provides evidence consistent with Werner's conjectured fractal dimension. The paper contains no fitted constants; the exponents and the high-dimensional correction factor N^{6-d} are explicit and parameter-free. The proof architecture is coherent, with the main probabilistic inputs cleanly separated into Proposition 3.2 and Lemmas 5.1–5.3. The most important caveat is that the quasi-multiplicativity bound (1.18) from the companion paper [5] is load-bearing for both main theorems; this is not circular, but it makes the present manuscript conditional on an external result.","major_comments":[{"comment":"The central input (1.18), quoted from the companion paper [5, Theorem 1.1], is not proved in this manuscript. It is used in two load-bearing places: it supplies the uniform contraction ratio in the Hopf argument of Proposition 3.2 (Case 2), and it enters the volume lower bound through the final line of (5.48) in Lemma 5.1. I do not regard this as circular, since (1.18) is a parameter-free two-point factorization whose statement does not mention IICs, but the dependence is structural rather than cosmetic. The authors should state the publication status of [5] and verify explicitly that the absorbing boundaries qD_k ∪ qD_{k+1} satisfy the hypotheses of (1.18) at every scale k, uniformly in the construction of Section 3.2.","section":"§3.3, Eqs. (3.95)–(3.100); §5.1, Eq. (5.48)"},{"comment":"The text says it completes the proof of 'Proposition 3.3', but the proposition proved in this section is Proposition 3.2; this should be corrected.","section":"§3.3, last paragraph"}],"minor_comments":[{"comment":"The phrase 'sigh cluster' should read 'sign cluster'.","section":"§1, after Eq. (1.3)"},{"comment":"Werner's conjecture is referred to as 'Werner (2016)' in the abstract and around Conjecture 1, while the reference [42] is dated 2021; the citation should be made internally consistent.","section":"§1, Conjecture 1 and Abstract"},{"comment":"The assertion that setting m = M yields n0 ≥ c_*(d,ε)M is terse: with the recurrence (3.65), n0 is related to M by a product depending on λ and K, so c_* must absorb that product. Please spell out this dependence or choose n0 first and then define m accordingly.","section":"§5.1, after Eq. (5.7)"},{"comment":"'Conseuqently' is a typo for 'Consequently'.","section":"§5.1, paragraph before Lemma 5.1"},{"comment":"The discussion explaining the bold-font superscripts N, x, T is more confusing than helpful; a simpler statement that these are labels, not numerical parameters, would suffice.","section":"§3.1, notation after Eq. (3.6)"}],"recommendation":"minor_revision","confidential_remarks":"The paper is well within the scope of the journal and the mathematical claims are substantial. The main editorial risk is the companion paper [5]: both main theorems are conditional on its quasi-multiplicativity estimate. I recommend that the editors verify that [5] is available and has been refereed before final acceptance, or require the authors to include a precise statement of (1.18) and its hypotheses in an appendix."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. First, this is the real thing: it proves existence and equivalence of four natural IIC constructions for critical GFF level sets and loop soups on the metric graph, for all d >= 3 except d = 6, and shows the incipient cluster is almost surely infinite and one-ended. Second, the whole proof is carried by a quasi-multiplicativity bound, (1.18), taken from the same authors' companion preprint [5]. The paper is honest about this, but if that bound fails, both main theorems fall.\n\nWhat's actually new: prior IIC literature covers Bernoulli percolation and FK-Ising; nothing existed for GFF level sets or loop soups. The adaptation of Basu-Sapozhnikov is nontrivial because long loops in the loop soup break the independence used in Bernoulli percolation. The paper handles that with a cluster-decomposition argument and a careful BKR-type inequality. The volume growth result (typical volume M^{(d/2+1) and 4} under conditioning on hitting the boundary) matches Werner's conjectured fractal dimension, and the anti-concentration theorem (Theorem 1.3) is a nice addition, indicating the scaling limit, if it exists, should be stochastic. Proofs are detailed, the structure is coherent, and there are no fitted constants or invented entities.\n\nSoft spots, in proportion. The main one is the external quasi-multiplicativity. It is load-bearing in two places: the Hopf contraction step in Section 3.3 needs the kernel in (3.75) to satisfy condition (3.95), which is asserted from (1.18) without a detailed check that the absorbing boundary qD_k union qD_{k+1} satisfies the hypotheses uniformly in k; and Lemma 5.1 uses (1.18) to get the volume lower bound. The paper also imports several auxiliary lemmas from [5] (Lemmas 2.3-2.6, 5.2, and others), so the external dependence is broader than a single theorem. This is not a red flag, since companion papers are normal, but it means a referee must read [5] before trusting Theorems 1.1 and 1.2. The d=6 exclusion is handled honestly in Remark 1.6, which gives partial results with poly-logarithmic errors.\n\nWho this is for: anyone working on GFF percolation, loop soups, or IIC theory. A serious referee should engage with it, with [5] in hand. I would want the referee to verify that (1.18) really applies to the kernels in the Hopf step, since that is the one place the paper is terse. Overall, this deserves peer review rather than desk rejection or immediate acceptance; I would expect substantial but fixable comments.","headline":"Strong paper: first IIC construction for GFF level sets and loop soups, with volume growth matching Werner's conjecture, but the proof leans heavily on a companion paper's quasi-multiplicativity that a referee must verify.","tokens_in":53180,"tokens_out":3541,"would_cite":true,"duration_ms":32186,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","60G60","82B43"],"pacs":[],"model":"deepseek-v4-flash","headline":"Four critical-cluster conditionings converge to one infinite, one-ended cluster for d≥3, d≠6, with volume M^min(d/2+1,4).","keywords":["incipient infinite cluster","Gaussian free field","metric graph","loop soup","quasi-multiplicativity","one-arm exponent","volume growth","fractal dimension"],"falsifier":"Choose a dimension $d\\ne6$ and an increasing cylinder event, and compute the four limiting conditional probabilities in (1.12), (1.13), (1.14), and (1.17) on that event; if any two limits differ, Theorem 1.1 is false. A cheaper check is the excluded case $d=6$: showing there that the four limits already fail to coincide, or that the conditioned volume has a power of $M$ different from $M^4$ up to polylogarithmic factors, would locate the boundary of the result.","tokens_in":52087,"feed_emoji":"∞","tokens_out":11143,"duration_ms":95456,"temperature":0.7,"pith_summary":"At criticality the cluster containing the origin is almost surely finite, so \"the cluster that manages to percolate\" cannot be obtained by a direct conditioning. The paper proves that four standard approximations of this impossible event — conditioning on connection to the boundary of a growing box, on being in an infinite cluster just above criticality, on connection to a lattice point tending to infinity, or on the capacity of the cluster exceeding a large threshold — each have a limit, and that the four limits are the same probability measure, for every dimension $d\\ge3$ except $d=6$. Under this common incipient infinite cluster law the critical cluster is almost surely infinite and one-ended. The same statement holds for the critical loop soup, by the isomorphism coupling with the Gaussian free field. If the result is right, it gives a canonical object on which to study the conjectured scaling limit of critical clusters, and the paper's volume estimate of order $M^{\\min(d/2+1,4)}$ matches the conjectured fractal dimension.","feed_headline":"One infinite cluster emerges from four critical-cluster conditionings","feed_subtitle":"All four limiting conditionings coincide: one infinite cluster, volume growing like M^(d/2+1) below 6 and M^4 above.","key_machinery":"The load-bearing input is quasi-multiplicativity, bound (1.18): for sets inside $B(cN^{1\\boxdot 2/(d-4)})$ and outside $B(CN^{1\\boxdot (d-4)/2})$, the probability of a connection through an intermediate box factorizes into the product of the two one-sided connection probabilities up to a constant and a correction factor $N^{0\\boxdot(6-d)}$. This factorization, proved in the companion paper and quoted as a black box, lets the authors decompose a large critical cluster into approximately independent annular pieces. The proof of uniform convergence then runs through a contraction theorem for positive integral operators, exactly as in the classical percolation framework the paper adapts, and the loop-soup IIC is transferred to the Gaussian free field by the isomorphism theorem identifying squared local times of the loop soup with the squared field.","core_discovery":"The central claim, Theorem 1.1, is that for any $d\\ge3$ with $d\\neq6$, the four limiting measures displayed in (1.12)--(1.14) and (1.17) exist and are equivalent. Under their common law $P_{d,\\mathrm{IIC}}$, the incipient infinite cluster $C^{\\ge0}$ is almost surely infinite and one-ended: for every $N$, $C^{\\ge0}\\setminus B(N)$ contains exactly one infinite cluster. Theorem 1.2 adds a quantitative self-similarity statement: conditioned on $\\{0\\leftrightarrow\\partial B(N)\\}$, the volume of $C^{\\ge0}\\cap B(M)$ lies between two constants times $M^{\\min(d/2+1,4)}$ with probability at least $1-\\epsilon$, provided $N\\gg M$; Theorem 1.3 shows this volume has a non-degenerate lower tail, so the scaling limit, if it exists, is not deterministic. Along the way the paper obtains, in Theorem 1.4, that under the same conditioning the connection probability to a point $y\\in\\partial B(M)$ is of order $M^{-[(d/2-1)\\boxdot(d-4)]}$.","pith_inferences":["A natural next test is the remaining dimension $d=6$: the paper's Remark 1.6 already obtains volume bounds with error exponent $\\varsigma(M)=\\ln\\ln M/\\sqrt{\\ln M}$, but the four IIC limits are not compared there; a polylogarithmic version of quasi-multiplicativity would likely be the missing ingredient.","Because Type (4) (capacity) and Type (1) (diameter) conditionings coincide, capacity behaves like a robust proxy for the size of the critical cluster; I conjecture that a volume-based conditioning, if a usable estimate for the volume tail becomes available, will also produce the same measure.","The framework suggests extensions to other transient graphs on which the GFF level-set has the same one-arm and quasi-multiplicativity estimates, where the same four-conditioning equivalence should hold.","The one-endedness plus volume growth exponent could feed random-walk and spectral-dimension questions on the IIC, in analogy with random-walk results on the high-dimensional percolation IIC."],"forward_implications":["The four IIC definitions in (1.12), (1.13), (1.14), and (1.17) can be used interchangeably for $d\\ge3$, $d\\neq6$; whichever conditioning is analytically most convenient is legitimate.","Under $P_{d,\\mathrm{IIC}}$, every truncation $C^{\\ge0}\\setminus B(N)$ contains a unique infinite component, so radial exploration of the incipient cluster is well defined.","Conditioned on the critical cluster reaching $\\partial B(N)$, its volume inside $B(M)$ is with high probability of order $M^{\\min(d/2+1,4)}$ for $N\\gg M$; that is the same exponent as the conjectured fractal dimension of the scaling limit.","The anti-concentration estimate (Theorem 1.3) implies that any scaling limit obtained from these IICs is stochastic rather than a deterministic fractal.","The normalized two-point function under the connection conditioning decays as $M^{-[(d/2-1)\\boxdot(d-4)]}$ (Corollary 1.5), providing a quantitative test for the IIC measure."],"supporting_citations":[{"why":"supplies the framework for constructing and proving equivalence of Types (1) and (2) IICs under a quasi-multiplicativity hypothesis.","marker":"[2]"},{"why":"proves the quasi-multiplicativity bound (1.18) for critical GFF level-sets that the present paper assumes as its main input.","marker":"[5]"},{"why":"gives the loop-soup/GFF isomorphism theorem used to transfer IICs and sign-cluster identities between the two models.","marker":"[29]"},{"why":"supplies the one-arm exponent $\\theta_d(N)\\asymp N^{-2}$ for $d>6$ and high-dimensional capacity and volume bounds used in the high-dimensional parts of Theorems 1.2 and 1.4.","marker":"[3]"},{"why":"supplies the low-dimension one-arm exponent $\\theta_d(N)\\asymp N^{-d/2+1}$ and annulus crossing estimates (1.7) used for $3\\le d\\le 5$.","marker":"[4]"},{"why":"gives the capacity decay (1.16) for critical clusters, the exact asymptotic needed for the Type (4) capacity IIC.","marker":"[11]"},{"why":"gives the supercritical percolation probability $P(0\\leftrightarrow\\infty)\\asymp h$ for level-sets at height $-h$, used in the Type (2) IIC and in Lemma 3.4.","marker":"[7]"},{"why":"states the conjectured scaling limit with fractal dimension $(d/2+1)\\boxdot 4$, which the volume-growth theorem targets.","marker":"[42]"},{"why":"supplies the contraction theorem for positive integral operators used to turn quasi-multiplicativity into exponentially fast convergence of the conditional probabilities.","marker":"[25]"}],"fun_headline_variants":["Four critical-cluster limits converge to one infinite cluster","Incipient infinite clusters unified for GFF and loop soups","Volume growth exponent confirms self-similar critical clusters","Critical cluster volume scales as M^4 or M^(d/2+1)","One-ended infinite cluster from four equivalent conditionings"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument takes as a black box the companion paper's quasi-multiplicativity bound: the probability that two far-apart sets are connected is within constants, times a correction factor $N^{0\\boxdot(6-d)}$, of the product of the probabilities that each set reaches the boundary of an intermediate box; if that factorization fails in some dimension, the existence and equivalence of the IICs and the volume estimates collapse.","fun_headline_variants_meta":{"raw":{"variants":["Four critical-cluster limits converge to one infinite cluster","Incipient infinite clusters unified for GFF and loop soups","Volume growth exponent confirms self-similar critical clusters","Critical cluster volume scales as M^4 or M^(d/2+1)","One-ended infinite cluster from four equivalent conditionings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000328,"raw_usage":{"total_tokens":1997,"prompt_tokens":1273,"completion_tokens":724,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":889,"completion_tokens_details":{"reasoning_tokens":642}},"tokens_in":889,"tokens_out":724,"duration_ms":6358,"temperature":1.0,"reasoning_tokens":642,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:26:15.482919+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose a dimension $d\\ne6$ and an increasing cylinder event, and compute the four limiting conditional probabilities in (1.12), (1.13), (1.14), and (1.17) on that event; if any two limits differ, Theorem 1.1 is false. A cheaper check is the excluded case $d=6$: showing there that the four limits already fail to coincide, or that the conditioned volume has a power of $M$ different from $M^4$ up to polylogarithmic factors, would locate the boundary of the result.","supporting_citations":[{"cited_title":"Basu and A","cited_arxiv_id":null,"evidence_quote":"supplies the framework for constructing and proving equivalence of Types (1) and (2) IICs under a quasi-multiplicativity hypothesis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the loop-soup/GFF isomorphism theorem used to transfer IICs and sign-cluster identities between the two models."},{"cited_title":"Cai and J","cited_arxiv_id":null,"evidence_quote":"supplies the one-arm exponent $\\theta_d(N)\\asymp N^{-2}$ for $d>6$ and high-dimensional capacity and volume bounds used in the high-dimensional parts of Theorems 1.2 and 1.4."},{"cited_title":"Drewitz, A","cited_arxiv_id":null,"evidence_quote":"gives the capacity decay (1.16) for critical clusters, the exact asymptotic needed for the Type (4) capacity IIC."},{"cited_title":"Ding and M","cited_arxiv_id":null,"evidence_quote":"gives the supercritical percolation probability $P(0\\leftrightarrow\\infty)\\asymp h$ for level-sets at height $-h$, used in the Type (2) IIC and in Lemma 3.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"states the conjectured scaling limit with fractal dimension $(d/2+1)\\boxdot 4$, which the volume-growth theorem targets."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the contraction theorem for positive integral operators used to turn quasi-multiplicativity into exponentially fast convergence of the conditional probabilities."}],"review_version":1}