arXiv++ Combinatorics

We study the optimal rectangular-discrepancy approximation of permutons by finite permutations. We transfer bounds from discrepancy theory to this more restricted setup. Moreover, we show that superlinear approximation can occur only for permutons supported by graphs of measure-preserving functions, and demonstrate how the local regularity of this function obstructs approximability. We also consider the biased Brownian separable permuton and prove a lower bound on its approximation error by showing that its supporting measure-preserving function has Lipschitz points almost surely.

In the past two years, several points of view have been proposed to address the question of the generalization of the theory of free probability to random tensors with different invariances, and it is unclear at this point whether they lead to the same notions of tensorial free cumulants and freeness. One way to approach this problem, developed by Collins, Gurau and the second named author for local unitary invariant random tensors, relies on finite size quantities involving averages over the invariance group, and whose asymptotics naturally possess the properties expected for tensorial generalizations of free cumulants of arbitrary orders. At this point, this approach has only been carried out for certain distributions, and for a subset of the moments that define such theories, and a more systematic and exhaustive study is lacking. This is the program initiated in this paper: we link this approach to the one proposed by Nechita and Park; extend a number of their results as well as those of the aforementioned paper to arbitrary orders of fluctuations, thereby generalizing higher order free cumulants; push further the study of distributions with larger invariance groups; detail the link with the asymptotics of the free-energies of the tensor HCIZ and BGW integrals; and provide formulae for tensorial free cumulants of products of tensors. Another important question is that of the definition of concrete distributions whose tensorial free-cumulants take non-trivial values. We compute the tensorial free cumulants for Gaussian random tensors with non-trivial covariances, and show that they provide such examples.

A set of integers is primitive if no number in the set divides another. We introduce a new method for bounding Erdős sums of primitive sets, suggested from output of GPT-5.4 Pro, based on Markov chains with von Mangoldt weights. The method leads to a host of applications, yet seems to have been overlooked by the prior literature since Erdős's seminal 1935 paper. As applications, we prove two 1966 conjectures of Erdős-Sárközy-Szemerédi, on primitive sets of large numbers (#1196) and on divisibility chains (#1217). The method also provides a short proof of the Erdős Primitive Set Conjecture (#164), as well as the related claim that 2 is an ''Erdős-strong'' prime. Moreover, the method resolves a revised form of the Banks-Martin conjecture, which has long been viewed as a unifying `master theorem' for the area.

We present two complementary proofs that, if the lengths of $n$ sticks are sampled at random, then the probability that no $p+1$ sticks can form a $(p+1)$-sided polygon can be expressed as the product of the reciprocals of a series of terms involving the $p$-step Fibonacci numbers. The first proof uses matrix algebra to extend the method previously used by Sudbury et al. to derive expressions for the probabilities of not being able to form triangles and quadrilaterals. The second alternative proof uses a different approach based on expressions for the minimum and maximum lengths of each stick that are compatible with the constraint of not being able to form a $(p+1)$-sided polygon, and provides insights into the structure of the probability expressions and the underlying reason that they include the Fibonacci numbers. Furthermore, the approach is developed in a generalised way that can, in principle, be applied to sticks randomly sampled from any probability distribution.

Denote by $p(k)$ the limit, as $n \rightarrow \infty$, of the probability that a random permutation on a set of size $n$ has an invariant set of size $k$. We give an asymptotic formula for $p(k)$, showing that it is asymptotically $f(\{\log_2 k\}) k^{-δ} (\log k)^{-3/2}$ where $δ= 1 - \frac{1 + \log \log 2}{\log 2} \approx 0.086$ and $f$ is a smooth, positive, function on $\mathbb{R}/\mathbb{Z}$, which we will describe explicitly. The function $f$ satisfies $\frac{\max f}{\min f} < 1 + 2 \times 10^{-7}$ and we conjecture that it is not constant. Estimating $p(k)$ is a model for the more well-known question which asks for an estimation of $M(n)$, the number of distinct elements in the $n$-by-$n$ multiplication table. By elaborating on the techniques in this paper, we will give an asymptotic for $M(n)$ in forthcoming work.

We study the coupon collector with deterministic expiration: one coupon is drawn at each time, and each coupon remains active for exactly $M$ draws. Completion occurs when all $n$ coupon types are simultaneously active. Equivalently, the current length-$M$ sliding window of draws must contain all $n$ types. The central object is not the one-time probability that a random window is onto, but the stationary flux of new entries into the onto-window set. We compute this flux exactly: \[ μ_{n,M} =\Pbb(W_{t-1}\text{ is not onto},\ W_t\text{ is onto}) =\frac{(n-1)(n-1)!S(M-1,n-1)}{n^M}, \] where $S(\cdot,\cdot)$ denotes a Stirling number of the second kind. Under a quantitative subcritical separation condition, satisfied in particular by every fixed integer scale $M=\floor{αn\log n}$, $0<α<1$, we prove local declumping and obtain \[ μ_{n,M}T_{n,M}\Rightarrow \Exp(1). \] For the fixed subcritical scale $M=\floor{αn\log n}$, $0<α<1$, this gives the logarithmic scale \[ \log T_{n,M}=n^{1-α}+o_{\mathbb P}(n^{1-α}), \qquad \log \Ebb T_{n,M}=n^{1-α}+o(n^{1-α}), \] and, when $α>1/2$, the sharper normalization \[ n^{-α}e^{-n^{1-α}}T_{n,M}\Rightarrow \Exp(1), \qquad \Ebb T_{n,M}\sim n^αe^{n^{1-α}}. \] Thus the leading scale proposed in the Math StackExchange discussion is made rigorous; the exact finite-$n$ flux gives the canonical normalization throughout the subcritical range. The result is a sliding-window companion to rare-void entry-flux methods for nonmonotone coupon collectors.

We develop a terminal-defect method for the double Dixie cup problem and use it to prove the finite-variance extremality conjecture of Doumas and Papanicolaou. For every \(m\ge1\) and \(N\ge2\), among all positive coupon probability vectors \(p=(p_1,\ldots,p_N)\), the variance of the time \(T_m(N)\) to collect \(m\) complete sets is uniquely minimized at the uniform vector. We prove the stronger radial statement that the variance is strictly increasing along every ray from the uniform vector. The proof is finite-\(N\) and exact: after Poissonization, the completion time is a maximum of independent Erlang variables, and the radial derivative of its distribution is compared to a size-biased law using a monotone-likelihood-ratio argument based on a log-scale monotonicity property of the Gamma reverse hazard. The same framework gives a growing-multiplicity Gumbel theorem in the equal-probability case, with expectation and variance asymptotics on the inverse gamma-tail scale. This recovers the fixed-\(m\) equal-probability variance asymptotic stated as Conjecture 1 by Doumas and Papanicolaou, classically known for \(m=1\), and extends the mechanism to \(m=m_N\). We also illustrate the unequal-probability theory with endpoint-Laplace limits for power-law probabilities.

Christoph, Draganić, Girão, Hurley, Michel, and Müyesser conjectured that, when $d\mid n$, the expected number of cycles in a uniformly random cycle-factor of a directed $d$-regular graph on $n$ vertices is uniquely maximised by the disjoint union of $n/d$ copies of the complete looped digraph $K_d^\circ$, with value $(n/d)H_d$ [FOCS 2025]. We disprove this conjecture in the strongest possible range. For every $d\ge 3$ and every multiple $n=kd$ with $k\ge 2$, we construct a directed $d$-regular graph on $n$ vertices whose uniformly random cycle-factor has expected cycle count strictly larger than $kH_d$. We also show that the conjectured extremal picture is correct in degree $d=2$, giving a sharp dichotomy between degree two and all higher degrees.

The continuous linear $P(k; N, w)$ and circular scan statistics $P_c(k; N, w)$ are fundamental tools in probability and spatial statistics, frequently used to detect clustering in uniform data. Let $X_1, X_2, \dots, X_N$ be independently and uniformly distributed random variables on a unit interval or unit ring. The exact distribution of these scan statistics relies on the minimum window width required to capture exactly $k$ points. Furthermore, the survival function $1 - P_c(k; N, w)$ directly corresponds to the geometric probability that if $N$ arcs of length $1 - w$ are uniformly and randomly placed on a unit circle, every point on the circle is covered at least $N + 1 - k$ times. Historically, evaluating the exact cumulative distribution functions, $P(k; N, w)$ and $P_c(k; N, w)$, relies heavily on complex recursive approximations. In this paper, we bypass these traditional recursive methods to derive direct, generalized closed-form expressions for some linear and circular continuous scan statistics. Specifically, we present the exact analytical solutions for $P_c(N - 1; N, w)$, $P_c(3; N, w)$, and $P(3; N, w)$ for arbitrary values of $N$ and window width $w$. These newly derived closed-form expressions not only provide exact baseline distributions for extreme spacings but also significantly simplify computational complexity compared to existing iterative approaches.

An $n$-tuple $D=(d(1),\dots,d(n))$ is a \emph{feasible degree sequence} if there is a graph on $\{1,\dots,n\}$ such that $i$ has degree $d(i)$. Any such graph will have $m=\sum_{i=1}^n d(i)/2$ edges. Letting $G(D)$ be a graph chosen uniformly from those with the given degree sequence, we upper-bound the probability that $G(D)$ is disconnected based on the number of vertices of degree $d$ for small $d$, and develop a powerful tool for proving such bounds. If there are any vertices of degree zero the probability $G$ is disconnected is $1$, so we assume there are no such vertices. Our results then imply that if there are $o(\sqrt{m})$ vertices of degree $1$ and $o(m)$ vertices of degree 2 then with high probability $G$ is connected, while if there are no vertices of degree 1 or 2 then the probability $G$ is disconnected is $O(\frac{n^4}{m^6})$.

We study a natural analogue of Ulam's problem for random rooted trees distributed according to a Plancherel-type measure. This probability measure is closely related to the classical Plancherel measure on integer partitions. For a Plancherel random tree $T_n$ with $n$ vertices, we investigate the asymptotic behavior of its height $H_n$, defined as the maximal distance from the root to a leaf. We prove that this height grows logarithmically. More precisely, there is a one-parameter family of random trees $(T_n(θ))_{n \in \mathbb{N}}$ indexed by $θ>0$ such that $\frac{H_n}{\log n}$ converges in probability to $c_\star(θ)$, where $c_\star(θ)$ is an explicit constant depending on the parameter $θ$. The case of Plancherel trees corresponds to the parameter $θ=2$. The proof is based on the fact that the Plancherel random trees can be viewed as Ewens fragmentation trees, for which the height exhibits a sharp threshold phenomenon. An upper bound is obtained via $s$-mass functionals and contraction estimates, while the lower bound is derived by embedding the model into a branching random walk with logarithmic displacements governed by a Poisson--Dirichlet distribution. The constant $c_\star(θ)$ is characterized through a variational principle associated with this branching random walk.

This is Part II of our project on block-weighted planar maps and Liouville quantum duality. Focusing on the scaling properties at the dual critical point, we derive the conditional distribution of the root block size given the total size, as well as, conversely, the distribution of the total size for a fixed root block size. We show that these laws are in perfect agreement with the results of Liouville quantum gravity (LQG), obtained by modifying the standard Liouville random measure with additional atomic contributions representing localized quantum areas. The ratio of dual and direct partition functions with punctures is shown to be universal, its explicit LQG expression exactly matching its combinatorial analogue. We also investigate the block distance profile for doubly rooted maps, which is here rigorously related to the distance profile of maps consisting of a single block. Finally, we analyze the multifractal properties of the usual and dual Liouville measures, predicting the associated spectra, from both quantum and Euclidean perpectives. We illustrate our results through specific realizations of block-weighted planar maps, i.e., quadrangulations decomposed into simple blocks, tree-like structures formed by attaching quartic maps, and bicubic maps decomposed into 3-connected blocks. For each model, we give the single non-universal constant which uniquely determines the strength of the corresponding atomic Liouville measure.

We prove the existence of the local limit of uniform random d-regular bipartite planar maps, for every $d\geq 3$, as the number of vertices tends to infinity. The proof relies on a bijection between maps and so-called blossoming trees established in a previous work. After proving local convergence of the associated decorated trees, we extend the bijection to infinite trees and transfer the convergence to planar maps. The limiting object is almost surely one-ended and recurrent for the simple random walk.

We study Mallows random permutations conditioned to avoid a given pattern $α$ of length~$3$. When the bias parameter is of the form $e^{β/n}$, we prove that these permutations converge to a non-trivial explicit deterministic permuton that depends on the pattern $α$ and on the parameter $β$. Along the way, we provide parametrizations for $α$-avoiding permutons, and establish a large deviation principle for uniform $α$-avoiding permutations. As a byproduct of the proof, we also obtain asymptotic estimates of two versions of $q$-Catalan numbers in the regime $q=e^{β/n}$.

We consider a recursive record-filtering procedure, which we informally call Disappear-Sort. Let $D_n$ denote the random variable giving the required number of passes in Disappear-Sort to eliminate a sequence of length $n$ sampled as i.i.d. copies of a continuous random variable $X$, where each pass retains the left-to-right records and discards all remaining entries. We show that this procedure admits two natural probabilistic interpretations. For the resampling variant we prove that $d_n=\mathbb{E}[D_n]$ satisfies an exact recurrence involving the unsigned Stirling numbers of the first kind. For the non-resampling variant, we associate to a permutation $p_n\in S_n$ a natural poset and prove that the recursive Disappear-Sort layers form an antichain decomposition of this poset. We deduce that the total number of passes equals $L(p_n)$, where $L(p_n)$ is the length of the longest decreasing subsequence of $p_n$. We then show that for a uniform random permutation of size $n$, the expectation $\mathbb{E}[D_n]$ of this second variant coincides with the expected first-column length of a Plancherel-random Young diagram. Using the Robinson--Schensted correspondence, we obtain an exact formula for this expectation in terms of partitions and standard Young tableaux, and classical Plancherel asymptotics then yield $\mathbb{E}[D_n]\sim 2\sqrt{n}$, with fluctuations on the $n^{1/6}$ scale governed by the Tracy--Widom law derived by Baik, Deift and Johansson. We conclude with an $O(n\log n)$ implementation.

The random transposition shuffle on repeated cards induces a Markov chain on the quotient space of arrangements with multiplicities, and is equivalent to the many-urn mean-field Bernoulli-Laplace model introduced by Scarabotti. Writing $n=ml$, where there are $m$ card types and each type appears $l$ times, we determine the limiting profile for the total variation distance to stationarity at times $t=\frac{n}{2}\left(\log n-\frac{1}{2}\log l+c\right)$, under the assumption $l=ω(1)$. Scarabotti previously established that this process exhibits cutoff at time $\frac{n}{2}(\log n-\frac{1}{2}\log l)$; our result refines this by identifying the precise asymptotic shape of convergence inside the cutoff window. We show that the limiting profile is asymptotically Gaussian, with different explicit forms in the regimes $m$ fixed and $m=ω(1)$. Together with our previous work on the fixed-$l$ regime, where the limiting profile is of Poisson type, this yields the cutoff profile for the random transposition shuffle on $n=ml$ repeated cards for the full range of parameters $m$ and $l$. Our argument has two main steps. First, we combine Scarabotti's Fourier-analytic framework for the many-urn Bernoulli-Laplace model with the approximation method of Jain-Sawhney (arXiv:2410.23944). More precisely, we compare the original shuffling measure with an explicitly tractable auxiliary measure directly on the repeated card quotient space, rather than passing through an intermediate comparison on the full symmetric group; this step relies in particular on our new estimates for Kostka numbers. Second, we reduce the limiting-profile problem to quotient fixed-point statistics and analyze them via Hoeffding-type combinatorial central limit theorems.

We study the expected number of rolls required for the cumulative sum of a fair six-sided die to first enter a prescribed target set $H\subset\mathbb{Z}_{\ge0}$. A one-variable dynamic-programming formulation is introduced that removes dependence on the roll count. Within this framework, the infinite process is truncated at a large cutoff $N$ and corrected by an analytically derived overshoot term that accounts for the rare event of exceeding $N$ before entering $H$. Explicit bounds on this residual yield a strict two-sided estimate of the truncation error. The method is numerically efficient, requiring constant memory and linear time in the cutoff. For the perfect-square target set $H=\{n^2:n\in\mathbb{N}\}$, all quantities are evaluated explicitly, yielding \[ \mathbb{E}[T]=7.07976423755110510389555305690818489468\ldots, \] provably correct to 1,017 decimal places. This constitutes the most precise result known to date and establishes a general framework for high-accuracy computation of discrete hitting times.

We analyze the asymptotic behavior and scaling limits of large random matrices rescaled via the Sinkhorn algorithm to match prescribed row and column margins. For a random matrix with independent sub-exponential entries, we show that its Sinkhorn rescaling concentrates around the rescaling of its mean matrix, both at the level of the Schrödinger potentials and as random measures on the unit square, with explicit non-asymptotic rates. As the dimensions grow, the rescaled random matrix converges to the continuous static Schrödinger bridge (SSB) determined by the limiting margins and reference density. Around this scaling limit we develop a fluctuation theory: bulk rigidity for the empirical spectral distribution of the associated sample covariance matrix, and a central limit theorem for the empirical Schrödinger potentials of the rescaled empirical mean. Our analysis is driven by a new quantitative stability theory for the SSB, developed in three forms: Lipschitz continuity in the Hellinger distance under perturbations of the reference measure (kernel stability); Hölder-$1/2$ continuity in the Hellinger distance under $L^1$ perturbations of the margins (margin stability); and $L^\infty$ stability of the discrete Schrödinger potentials under margin perturbation (potential stability). Translated to the discrete random-matrix setting, these bounds yield the concentration and scaling-limit results, while a local law for random Gram matrices with a non-uniform variance profile drives the bulk rigidity. Our SSB stability theory may be of independent interest.

We study the edge-expansion of the graph of a random $0/1$ polytope $P^d_p$, defined as the convex hull of a random subset of the points in $\{0,1\}^d$ where every point is retained independently and with probability $p$. This problem was introduced more than twenty years ago in a work of Gillmann and Kaibel, and has been extensively studied ever since. We prove that, for every fixed $\varepsilon>0$ and every $p\in(0,1-\varepsilon]$, with high probability the graph of $P^d_p$ has edge-expansion $Θ(d)$. This improves the previously best known bound due to Ferber, Krivelevich, Sales and Samotij, and verifies, in a strong form, the celebrated Mihail-Vazirani conjecture for random $0/1$ polytopes. Although the expansion factor $Θ(d)$ is typically best possible for $p\ge 1/2+\varepsilon$, we also show that the behaviour changes drastically at $p=1/2$. Namely, for every fixed $\varepsilon>0$ and every integer $k\ge 2$, if $p\le 1/2-\varepsilon$, then with high probability the graph of $P^d_p$ has edge-expansion $Ω(d^k)$. Thus, random $0/1$ polytopes exhibit an interesting phase transition at $p=1/2$.

We study the community detection problem in the non-uniform hypergraph stochastic block model (HSBM), where hyperedges of varying sizes coexist. This setting captures higher-order and multi-view interactions and raises a fundamental question: can multiple uniform hypergraph layers below the detection threshold be combined to enable weak recovery? We answer this question by establishing a Kesten--Stigum-type bound for weak recovery in a general class of non-uniform HSBMs with $r$ blocks, generated according to multiple symmetric probability tensors. In the case $r=2$, we show that weak recovery is possible whenever the sum of the signal-to-noise ratios across all uniform hypergraph layers exceeds one, thereby confirming the positive part of a conjecture in (Chodrow et al., 2023). Moreover, we provide a polynomial-time spectral algorithm that achieves this threshold via an optimally weighted non-backtracking operator. For the unweighted non-backtracking matrix, our spectral method attains a different algorithmic threshold, also conjectured in (Chodrow et al., 2023). Our approach develops a spectral theory for weighted non-backtracking operators on non-uniform hypergraphs, including a precise characterization of outlier eigenvalues and eigenvector overlaps. We introduce a novel Ihara--Bass formula tailored to weighted non-uniform hypergraphs, which yields an efficient low-dimensional representation and leads to a provable spectral reconstruction algorithm. Taken together, these results provide a principled and computationally efficient approach to clustering in non-uniform hypergraphs, and highlight the role of optimal weighting in aggregating heterogeneous higher-order interactions.