{"id":"01d2cc38-ea94-4c79-8abc-c3b684a2bdc7","arxiv_id":"2412.19730","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper develops a d-dimensional permuton framework and proves that uniform Schnyder wood and d-separable permutations converge to explicit random high-dimensional permutons connected to SLE and LQG.","lead":"This paper builds a theory of high-dimensional permutons, the probability measures that describe the large-scale shape of high-dimensional permutations, and proves two natural families of random permutations converge to explicit random limit objects. The limits connect permutation combinatorics to Brownian trees, Schramm-Loewner evolutions, and Liouville quantum gravity surfaces, which is why a generalist might care.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 4.19's absolute-continuity transfer is the load-bearing step for Theorem 1.9 and is left as a sketch; the measurability/continuity of the coalescent-walk functionals in the driving walk is not proved, so pattern identification could fail.","rationale":"The reader's weakest assumption identifies the absolute-continuity/measurability transfer in Corollary 4.19 as the load-bearing step for Theorem 1.9, and I agree. The proof of Theorem 1.9 proceeds by showing convergence of pattern frequencies via Corollary 4.20, which depends entirely on Corollary 4.19 to pass from unconditioned coalescent-walk limits to the conditioned setting. The manuscript's proof of Corollary 4.19 is explicitly a sketch, and the crucial claim that the coalescent-walk processes are measurable, essentially continuous, functionals of the driving walks in the scaling limit is asserted without proof. If this transfer fails, even on a measure-zero-looking but positive-conditioned-measure set, the identification of the limiting pattern law with the skew-Brownian-permuton-driven distribution in Step 5 of Theorem 1.9 would collapse. I find no other concern that is comparably load-bearing: the tree-convergence input for Theorem 1.14 is a known external result, and the Proposition 5.7 typo (2^d instead of 2^{d-1}) does not affect the uniformity argument. The concern does not amount to a demonstrated contradiction or an internal inconsistency; it is a missing proof for a standard but delicate technique, so the appropriate verdict remains CONDITIONAL as the reader had it. The proposed concrete test—an independent h-transform derivation of Corollary 4.19 with explicit continuity checks—would settle whether the concern lands.","tokens_in":63260,"tokens_out":11150,"duration_ms":116387,"concrete_test":"Re-derive Corollary 4.19 for a fixed u ∈ (0,1) without the skipped technical steps: construct the conditioned walk from the unconditioned one via the h-transform and verify that the finite-dimensional distributions of (W_n, the red process at the ⌈nu⌉-th starting point, the green process at the mirrored starting point) converge to the SDE-driven limit, uniformly over u in a compact subinterval of (0,1). In particular, check that the SDE solution map (X,Y) ↦ Z^{(u)} in Eq. (15) is continuous in the sup-norm topology on the driving path on a set of full Brownian-excursion measure, and that the discrete maps converge to it in probability after the conditioning. If continuity fails, exhibit an explicit discontinuity set with positive conditioned-law measure; if it holds, the transfer is justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.9 is proved via pattern convergence, reduced in §4.3 to Corollary 4.20, which relies on Corollary 4.19. That corollary transfers the unconditioned joint convergence of Proposition 4.9 to the conditioned walk (staying in {x ≥ 0, y ≥ −1} and ending at the origin) by absolute continuity between Brownian motion and Brownian excursion on [ε, 1−ε], appealing to the assertion that 'measurability of the coalescent-walk processes in terms of the driving walks carries over from the discrete case to the scaling limit.' The proof is explicitly a sketch, deferring to [Bor22, Appendix E], and the assertion is not demonstrated. The delicate point is that the coalescent-walk paths are read at the random starting points jn (the ⌈nu⌉-th leftmost starting point, and its mirrored counterpart), which are functionals of the conditioned walk itself. If the discrete functionals W ↦ Z^{(j_n(u))} do not converge, in the appropriate topology, to the SDE solution map of Eq. (15) on a set of full measure under the conditioned law, then the limiting signs in Eqs. (35)–(36) could differ from those of the Brownian-excursion-driven SDE, breaking the identification ρ_k = P_μS[k] in Step 5 of the proof. The manuscript does not supply the required uniform-in-u or continuity estimates; it only gives the sketch and a reference. This is the least secure premise on which the main Schnyder wood limit rests.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of d-dimensional permutons for all d ≥ 2, proving that weak convergence of random d-permuton measures is characterized by convergence of all pattern frequencies, in expectation or in distribution (Theorem 1.5). It then applies this theory to two natural families: uniform 3-dimensional Schnyder wood permutations, which are shown to converge to a random 3-permuton built from two coupled skew Brownian permutons with parameters ρ = −√2/2 and q = 1/(1+√2) (Theorem 1.9), and uniform d-separable permutations, which are shown to converge to the Brownian separable d-permuton with all parameters 1/2 (Theorem 1.14). The proofs use bijections to two-dimensional walks and coalescent-walk processes for the Schnyder wood case, and bijections to conditioned Galton–Watson trees for the separable case.","tokens_in":63534,"tokens_out":6886,"duration_ms":75331,"significance":"This is a substantial contribution if the main results hold. The paper provides the first general high-dimensional permuton framework and identifies explicit random limiting permutons for two combinatorially meaningful classes of high-dimensional permutations. The limiting objects are connected to skew Brownian permutons, SLE-decorated LQG surfaces, and the continuum random tree, giving new evidence for universality of these 2D permuton families. The proof of Theorem 1.5 is detailed and self-contained, and the combinatorial bijections (Proposition 4.1, Proposition 5.4) are nontrivial and well motivated. The paper also contains explicit calculations, such as the expected inversion frequency in Remark 1.16, and it is honest about places where proofs are abbreviated or deferred to previous works.","major_comments":[{"comment":"Corollary 4.19 is the load-bearing step in the proof of Theorem 1.9, but its proof is explicitly a sketch. The argument transfers unconditioned coalescent-walk convergence (Proposition 4.9) to the conditioned walk by absolute continuity between Brownian motion and Brownian excursion on [ε, 1−ε], and then asserts that 'measurability of the coalescent-walk processes in terms of the driving walks carries over from the discrete case to the scaling limit.' This is not demonstrated. The delicate point is that the paths are read at the random starting points j_n and 2n−j_n, which are functionals of the conditioned walk itself, and the identification of the limiting signs in Eqs. (35)–(36) requires convergence of the discrete maps W ↦ Z^{(j_n(u))} to the SDE solution map of Eq. (15) in a topology that handles the random starting point. A reference to [Bor22, Appendix E] is not sufficient for a step on which the characterization of ρ_k = P_{μ_S}[k] rests. I ask the authors to provide a complete proof of Corollary 4.19, including the required continuity or uniform-in-u estimates, or to restructure the argument so that this conditioning transfer is proved directly.","section":"§4.3, Corollary 4.19"},{"comment":"The proof of Theorem 1.14 depends on an unproved structural assertion about the random induced subtree T_{n,I}: that, with high probability, T_{n,I} is a uniform binary plane tree with k leaves plus an extra root vertex, that the relevant internal vertices of T_n are pairwise at distance at least n^{1/4}, and that the parities of the numbers of 1s in the intervening swap sequences are asymptotically independent and uniform. The manuscript says this is 'well-known' and cites [Ald91a] and [BBFS20, Section 4], but the precise form needed here—especially the independence of the parities along long paths in a conditioned Galton–Watson tree—is not stated as a lemma, nor is it derived. Since this is the mechanism that makes the limiting signs uniform and independent in Proposition 5.8 for the case p_i = 1/2, this is another load-bearing point in a main theorem and should be proved or stated with a fully matching reference.","section":"§5.2, Theorem 1.14 proof"}],"minor_comments":[{"comment":"In the proof of Lemma 4.10, the sentence 'which occurs if and only if n − j is a starting point in WCg(W′)' should read '2n − j' rather than 'n − j'.","section":"§4.1, Lemma 4.10"},{"comment":"In Definition 4.12, the infinite red coalescent-walk process is denoted Z^{*,r}_n = WCr(W^{*,r}); the subscript n is a typo and should be omitted, since this process is defined on the infinite interval [0,∞).","section":"§4.2.1, Definition 4.12"},{"comment":"In the proof of Proposition 4.17, the truncation of the Riemann sum in Eq. (26) by dropping the first and last n^{1/4} terms is stated with an O(n^{-1/4}) error, but the justification is only sketched via the asymptotic bounds P(Z_k = 0) = O(k^{-1/2}). A short explicit estimate would make this easier to verify.","section":"§4.2.2, Proposition 4.17"},{"comment":"The limiting coalescent-walk paths are defined on [0,1], so the target space C((0,1),R)^2 in Corollary 4.19 is unnecessarily weak; stating convergence in C([0,1],R)^2 would be cleaner and would match the subsequent use of the paths at all times.","section":"§4.3, Corollaries 4.19–4.20"},{"comment":"The notation μ_S = (μ^g_S, μ^r_S) is informal: μ_S is a measure on [0,1]^3, not literally a pair of measures. The intended meaning is clear from Eq. (6), but a short clarifying sentence would avoid confusion.","section":"§1.2.1, Eq. (7)"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically rich and the main framework is convincing. My concerns are not about the overall validity of the strategy but about the completeness of two arguments on which the two main limit theorems rest: the conditioning transfer in Corollary 4.19 and the induced-subtree asymptotics in the proof of Theorem 1.14. Both are standard-looking and likely correct, but they are load-bearing and are currently either sketched or asserted with references. If the authors supply complete proofs, the paper would be suitable for publication. I would not recommend rejection based on the current version, because I did not find a definite error in the core derivations."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things. First, the d-dimensional permuton machinery in Section 2 is a clean generalization of the 2D theory, and the pattern-frequency characterization (Theorem 1.5) is proved carefully; that part deserves to become a standard reference. Second, the two limit theorems are genuinely new and largely convincing: uniform Schnyder wood permutations have a random 3D permuton limit built from two coupled skew Brownian permutons, and uniform d-separable permutations converge to the Brownian separable d-permuton. The Schnyder wood result in particular is a real step beyond existing 2D permuton work, and the connection to SLE/LQG objects is not just decoration.\n\nWhat the paper does well: it states the limits precisely, proves Theorem 1.5 in detail, and the two applications are supported by long, structured arguments involving bijections, coalescent-walk scaling, and Galton-Watson trees. The bijective encoding of Schnyder wood triangulations by a pair of coalescent-walk processes driven by one walk is elegant. I found no outright mathematical error in the parts I checked.\n\nSoft spots, in order of importance. The conditioning transfer in Corollary 4.19 is the one spot that needs referee attention. The proof is labeled a sketch and defers to [Bor22, Appendix E]; the claim that measurability of the coalescent-walk processes in terms of the driving walks carries over from the discrete case to the scaling limit is not demonstrated. If the discrete-to-continuum map W ↦ Z^{(j_n(u))} does not converge uniformly in u, the sign identifications in Step 3 of Theorem 1.9 could fail. My read is that this is a repairable gap rather than a false statement—the surrounding estimates in Lemma 4.18 and Proposition 4.17 make the intended argument plausible—but the manuscript as written does not complete it.\n\nMinor issues: Proposition 5.7 says the root is labeled by one of the 2^d sign sequences; it should be 2^{d-1}. The Galton-Watson-to-uniform-tree step in Theorem 1.14 is cited as well-known rather than proved; a referee should verify the cited [BBFS20] result covers exactly this conditioned subtree. Remark 1.15 promises a proof that it omits; harmless, but it should be marked as a claim.\n\nVerdict: the central argument holds up and the contributions are real. This deserves serious refereeing. If the gap in Corollary 4.19 is fixed or properly cited, this is a strong paper.","headline":"A clean d-dimensional permuton theory plus two genuinely new limit theorems, with one under-proved conditioning transfer in the Schnyder wood proof that needs referee attention.","tokens_in":64095,"tokens_out":4163,"would_cite":true,"duration_ms":41973,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60F17","05A05","60C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper introduces d-dimensional permutons—probability measures on [0,1]^d with uniform one-dimensional marginals—and proves that, as in the two-dimensional case, weak convergence of random d-dimensional permutations is equivalent to…","keywords":["d-dimensional permutons","Schnyder wood permutations","d-separable permutations","skew Brownian permutons","Brownian separable d-permuton","coalescent-walk processes","pattern frequencies","random geometry"],"falsifier":"Compute the expected frequency of the inversion pattern $(2,1)$ in the green marginal of a large uniform Schnyder wood permutation: the paper predicts $2/3$ because the skew Brownian permuton parameter gives $E[\\mathrm{freq}((2,1),\\mu^g_{\\rho,q})] = 1/3$ and the reflection in Eq. (7) flips it to $2/3$. A Monte Carlo estimate that persistently deviates from $2/3$ as $n$ grows would falsify Theorem 1.9. Alternatively, simulate the limiting coalescent-walk processes driven by a Brownian excursion of correlation $-\\sqrt{2}/2$ and compare the induced pattern distribution with $P_{\\mu_S}[k]$ for small $k$.","tokens_in":63009,"feed_emoji":"📐","tokens_out":4193,"duration_ms":42775,"temperature":0.7,"pith_summary":"This paper builds a theory of d-dimensional permutons and shows that weak convergence of random d-dimensional permutations is controlled by convergence of all pattern frequencies in expectation, exactly as in the classical two-dimensional setting. It then determines the random scaling limits of two natural families of high-dimensional permutations. Uniform 3-dimensional Schnyder wood permutations, which bijectively encode planar triangulations decorated by three spanning trees, converge to the Schnyder wood permuton, built from skew Brownian permutons with correlation $\\rho = -\\sqrt{2}/2$ and skewness $q = 1/(1+\\sqrt{2})$. Uniform $d$-separable permutations converge to the Brownian separable $d$-permuton with all parameters equal to $1/2$. These are explicit, random, fractal limits tied to objects from random geometry such as the continuum random tree, Schramm--Loewner evolutions, and Liouville quantum gravity surfaces.","feed_headline":"High-dimensional permutons get explicit random limits","feed_subtitle":"Schnyder wood and d-separable permutations both converge to named Brownian-geometry permutons.","key_machinery":"The carrying objects are coalescent-walk processes: collections of one-dimensional random walks that start at specified points, never cross, and stick together once they meet, together with the total orders they induce on their starting points. For Schnyder wood permutations, a bijective chain links triangulations to strings, then to a two-dimensional walk, then to a green and a red coalescent-walk process driven by the same walk in opposite time directions; patterns of the permutation are read from the signs of sample paths at random starting points. The scaling limit of these processes is identified through the SDE system defining skew Brownian permutons, with the conditioning transfer handled by absolute-continuity arguments. For $d$-separable permutations, the bijection with sign trees and swap trees, combined with conditioned Galton--Watson trees whose asymptotic shape is binary, yields the Brownian excursion construction of the Brownian separable $d$-permuton.","core_discovery":"The central discovery is that a coherent high-dimensional permuton theory exists and has nontrivial, explicit examples. Theorem 1.5 characterizes convergence of random $d$-dimensional permutations by convergence of expected pattern frequencies. Theorem 1.9 identifies the Schnyder wood permuton $\\mu_S$ as the limit of uniform Schnyder wood permutations: it is the Lebesgue measure pushed through $(t, \\phi^g_{\\rho,q}(1-t), 1-\\phi^r_{\\rho,q}(t))$, where $\\phi^g$ and $\\phi^r$ come from coupled skew Brownian permutons driven by a Brownian excursion of correlation $\\rho = -\\sqrt{2}/2$ and its time-reversal, with $q = 1/(1+\\sqrt{2})$. Theorem 1.14 shows that uniform $d$-separable permutations converge to the Brownian separable $d$-permuton $\\mu^B_{1/2,\\dots,1/2}$, built from a single Brownian excursion whose local minima carry independent uniform sign vectors.","pith_inferences":["The time-reversal coupling between the two marginals suggests that other decorated planar map encodings might produce high-dimensional permutons whose coordinates are all functions of a single driving excursion; this is an editorial extrapolation, not stated in the paper.","The Brownian separable $d$-permuton with parameters $1/2$ is a plausible limit for other substitution-closed classes of high-dimensional permutations, with class-specific parameters; the paper states this as a conjecture.","A testable extension is to estimate, for moderately large $n$, the joint distribution of green and red pattern frequencies in uniform Schnyder wood permutations and compare it with the coupled skew Brownian permuton prediction, which would probe the strength of the dependence between the two marginals.","The theory likely applies to other high-dimensional permutation models defined via bijections with walks, where a finite set of coalescent-walk processes driven by a single walk could each encode one coordinate of the limiting permuton."],"forward_implications":["Uniform 3-dimensional Schnyder wood permutations converge in distribution to an explicit random 3-permuton, so the bijective encoding of triangulations has a continuum counterpart at the permuton level.","The two 2-dimensional marginals of the Schnyder wood permuton are skew Brownian permutons of known parameters, and the full 3-permuton is determined by either one of those marginals.","Uniform $d$-separable permutations converge to the Brownian separable $d$-permuton with all parameters $1/2$, giving an honest $d$-dimensional random limit whose lower-dimensional marginals do not determine the whole object.","The high-dimensional permuton theory reduces convergence problems to the computation of expected pattern frequencies, so future families of high-dimensional permutations can be handled by enumerative estimates.","The results extend known universality of skew Brownian permutons and Brownian separable permutons to higher-dimensional permutation classes.","The Schnyder wood permuton admits a geometric description in which the two marginals are obtained by comparing space-filling SLE curves of angles $0$, $2\\pi/3$, and $4\\pi/3$, linking the limit to SLE-decorated Liouville quantum gravity surfaces."],"supporting_citations":[{"why":"Provides the two-dimensional characterization of permuton convergence via expected pattern frequencies that Theorem 1.5 generalizes.","marker":"[BBF+20]"},{"why":"Introduces coalescent-walk processes and their use in proving convergence to skew Brownian permutons, which the Schnyder wood proof extends to a pair of processes.","marker":"[BM22b]"},{"why":"Defines the skew Brownian permutons and establishes existence and uniqueness of the SDE solutions used to describe the Schnyder wood limit.","marker":"[Bor23]"},{"why":"Supplies the convergence of Schnyder wood triangulations to SLE-decorated LQG surfaces and the value $q = 1/(1+\\sqrt{2})$.","marker":"[LSW24]"},{"why":"Gives the result that a skew Brownian permuton determines its driving Brownian excursion, used to prove that the Schnyder wood permuton is determined by one marginal.","marker":"[BG24]"},{"why":"Introduces $d$-separable permutations and their block-sum characterization, which the paper uses to build sign and swap trees.","marker":"[AM10]"},{"why":"Provides the Brownian separable permuton limit and the subtree sampling results used for the $d$-separable argument.","marker":"[BBFS20]"},{"why":"Gives the reconstruction of a Schnyder wood from its green tree via prefix flips, used to prove the bijection between Schnyder wood triangulations and permutations.","marker":"[Bon05]"}],"fun_headline_variants":["Random limits for high-dim permutons","Schnyder and d-separable perms get random limits","Theory identifies random high-dim permuton limits","High-dim permuton limits from Brownian geometry","Explicit random limits in high-dimensional permutons"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the unconditioned coalescent-walk convergence transfers to the conditioned setting uniformly over the random starting points used to read off patterns; if that conditioning transfer fails, identifying the limiting law of pattern frequencies in Theorem 1.9 breaks.","fun_headline_variants_meta":{"raw":{"variants":["Random limits for high-dim permutons","Schnyder and d-separable perms get random limits","Theory identifies random high-dim permuton limits","High-dim permuton limits from Brownian geometry","Explicit random limits in high-dimensional permutons"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000741,"raw_usage":{"total_tokens":3343,"prompt_tokens":1019,"completion_tokens":2324,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":2258}},"tokens_in":635,"tokens_out":2324,"duration_ms":19434,"temperature":1.0,"reasoning_tokens":2258,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:55:49.699197+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the expected frequency of the inversion pattern $(2,1)$ in the green marginal of a large uniform Schnyder wood permutation: the paper predicts $2/3$ because the skew Brownian permuton parameter gives $E[\\mathrm{freq}((2,1),\\mu^g_{\\rho,q})] = 1/3$ and the reflection in Eq. (7) flips it to $2/3$. A Monte Carlo estimate that persistently deviates from $2/3$ as $n$ grows would falsify Theorem 1.9. Alternatively, simulate the limiting coalescent-walk processes driven by a Brownian excursion of correlation $-\\sqrt{2}/2$ and compare the induced pattern distribution with $P_{\\mu_S}[k]$ for small $k$.","supporting_citations":[],"review_version":1}