Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Mathematics

math Watch Mathematics (sign in to use)

Thursday 9 April

Number Theory

A Dynamical Lifting Problem For Additive Polynomials

We introduce a dynamical analogue of the lifting problem for Galois covers of algebraic curves and find a negative solution for the collection of separable additive and semi-additive polynomials over $\overline{\mathbb{F}}_p$. We also explicitly compute the dimension of the space of linear conjugacy classes in $\mathcal{M}_{p^m}(\overline{\mathbb{F}}_p)$ which contain additive or semi-additive polynomials providing explicit counterexamples to Thurston Rigidity in positive characteristic.
Read the full abstract

Combinatorics

Schuttes property for sets of tournaments and an application to dice games

A tournament has Schuttes property $S_k$ if for every set of $k$ vertices, there is a vertex which dominates the set. In 1963, Erdos provided bounds for $f(k)$, the smallest order of an $S_k$ tournament. Schuttes property has various applications, including the design of unfair dice games. A set of dice introduced by James Grime motivates a generalization of Schuttes property to sets of tournaments: a set of tournaments on the same vertex set has property $S_k$ if for every set of $k$ vertices, there is a vertex which dominates the set in at least one of the tournaments. We explore this generalization and provide bounds on the fewest number of vertices needed to have an $S_k$ set of $m$ tournaments. We then apply these results to introduce a few new sets of dice similar to Grimes dice that can be used to play a game that gives one player an advantage.
Read the full abstract

History and Overview of Mathematics

Open Preparation, Human Explanation, and Instructor Synthesis: A Human-Scale Methodology for AI-Rich Higher Education

In AI-rich higher education, polished written mathematics has become easier to produce than trustworthy evidence of understanding. This article develops a human-scale methodology for service mathematics, with informatics as its main running case. Its central move is not prohibition of tools but relocation of evidential trust. Students prepare openly, often with digital assistance, but grade-relevant evidence shifts toward live explanation, contingent questioning, and cumulative observation against course outcomes. The design is guided by Realistic Mathematics Education, question-first task construction, short human-scale mathematical tasks, and instructor synthesis after student attempt. It contributes a weekly operational cycle, a realism framework distinguishing professional, disciplinary, and experiential realism, a middle-out white-box / black-box stance on tools, a bounded role for retrieval-grounded AI assistants for students and teachers, and a cumulative oral-evidence model for small and medium cohorts. The paper also provides concrete implementation artifacts: process figures, an ecology of problem types, time-budget estimates, an evidence hierarchy, and a five-grade oral rubric. This is a methodology paper rather than an effectiveness study. Its claim is that the proposed design is pedagogically coherent, operationally plausible for human-scale teaching settings, and responsive to current concerns about AI, oral evidencing, and active learning in undergraduate mathematics education.
Read the full abstract

Algebraic Geometry

Neutral representations in dimension $\leq 3$ and fields of moduli

A representation $V$ of an algebraic group $G$ induces a vector bundle $[V/G] \to BG$. The representation $V$ of $G$ is neutral if, for every twisted form $\mathcal{V} \to \mathcal{G}$ of $[V/G] \to BG$ over a field $k$, we have $\mathcal{G}(k) \neq \emptyset$. Twisted forms of representations arise in many ways, for instance as cohomology of families of varieties on residual gerbes of moduli spaces, and from quotient singularities. Moreover, every Tannakian category is the category of vector bundles on some gerbe. Because of this, studying neutral representations yields numerous applications, especially to problems about fields of moduli. The present article has three main results. First, we completely classify neutral, faithful representations of finite groups in dimension $\leq 3$. Second, we give a very general, computation-friendly result for proving that representations of finite abelian groups are neutral, in arbitrary dimensions. Third, we develop the abstract concept of the normalizer $\mathcal{G} \to \mathcal{N} \to \mathcal{H}$ of a morphism of gerbes $\mathcal{G} \to \mathcal{H}$ on an arbitrary site (twisted representations correspond to morphisms of gerbes $\mathcal{G} \to B\mathrm{GL}_{n}$), and show that the normalizer $\mathcal{N}$ only depends on the geometric type of $\mathcal{G} \to \mathcal{H}$.
Read the full abstract

Logic

Free Left Distributive Algebras and a Canonical Extension

Assuming a large cardinal hypothesis, Laver gave a representation of the monogenerated free left distributive algebra (LDA) using elementary embeddings and used this representation to prove many algebraic results. Some of these results were later proved by Dehornoy in ZFC, without the large cardinal hypotheses. However, there is an important algebraic result whose consistency strength is unknown. (See Laver (1995) and Dougherty & Jech (1997).) Recent results [arXiv:2508.02244] extend the connection between elementary embeddings of set theory and free LDAs to the many-generated case. Assuming large cardinals, we prove two results. First, we prove that finitely-generated free LDAs with distinct numbers of generators are $\Sigma_1$-elementarily equivalent but not $\Sigma_2$-elementarily equivalent. We also prove a partial structural analogue to Laver's representation of LDAs. We construct an extension of the monogenerated free LDA where application by any fixed element is an elementary embedding of LDAs. We argue that this extension is canonical by demonstrating homogeneity and universality properties. These results also provide additional examples of algebraic properties provable from large cardinals without known proofs from the standard axioms of set theory.
Read the full abstract

All authors Scott Cramer, Meng-Che "Turbo" Ho, Sheila K. Miller Edwards, Nam Trang

2604.08768 · 9 Apr · 36 pages. Cited in abstract: Richard Laver (1995). On the algebra of elementary embeddings of a rank into itself in Advances in Mathematics, vol. 110, pages 334-346; Randall Dougherty and Thomas Jech, (1997). Finite left-distributive algebras and embedding algebras, Advances in Mathematics, vol. 130, pages 201-241; and arXiv:2508.02244

Differential Geometry · Cross-listed from cs.CL

Revisiting Anisotropy in Language Transformers: The Geometry of Learning Dynamics

Since their introduction, Transformer architectures have dominated Natural Language Processing (NLP). However, recent research has highlighted an inherent anisotropy phenomenon in these models, presenting a significant challenge to their geometric interpretation. Previous theoretical studies on this phenomenon are rarely grounded in the underlying representation geometry. In this paper, we extend them by deriving geometric arguments for how frequency-biased sampling attenuates curvature visibility and why training preferentially amplify tangent directions. Empirically, we then use concept-based mechanistic interpretability during training, rather than only post hoc, to fit activation-derived low-rank tangent proxies and test them against ordinary backpropagated true gradients. Across encoder-style and decoder-style language models, we find that these activation-derived directions capture both unusually large gradient energy and a substantially larger share of gradient anisotropy than matched-rank normal controls, providing strong empirical support for a tangent-aligned account of anisotropy.
Read the full abstract

All authors Raphael Bernas, Fanny Jourdan, Antonin Poché, Céline Hudelot

Numerical Analysis · Cross-listed from quant-ph

Weak Adversarial Neural Pushforward Method for the Wigner Transport Equation

We extend the Weak Adversarial Neural Pushforward Method to the Wigner transport equation governing the phase-space dynamics of quantum systems. The central contribution is a structural observation: integrating the nonlocal pseudo-differential potential operator against plane-wave test functions produces a Dirac delta that exactly inverts the Fourier transform defining the Wigner potential kernel, reducing the operator to a pointwise finite difference of the potential at two shifted arguments. This holds in arbitrary dimension, requires no truncation of the Moyal series, and treats the potential as a black-box function oracle with no derivative information. To handle the negativity of the Wigner quasi-probability distribution, we introduce a signed pushforward architecture that decomposes the solution into two non-negative phase-space distributions mixed with a learnable weight. The resulting method inherits the mesh-free, Jacobian-free, and scalable properties of the original framework while extending it to the quantum setting.
Read the full abstract

Information Theory

Optimal Multi-bit Generative Watermarking Schemes Under Worst-Case False-Alarm Constraints

This paper considers the problem of multi-bit generative watermarking for large language models under a worst-case false-alarm constraint. Prior work established a lower bound on the achievable miss-detection probability in the finite-token regime and proposed a scheme claimed to achieve this bound. We show, however, that the proposed scheme is in fact suboptimal. We then develop two new encoding-decoding constructions that attain the previously established lower bound, thereby completely characterizing the optimal multi-bit watermarking performance. Our approach formulates the watermark design problem as a linear program and derives the structural conditions under which optimality can be achieved. In addition, we identify the failure mechanism of the previous construction and compare the tradeoffs between the two proposed schemes.
Read the full abstract

Dynamical Systems

Effective equidistribution of unipotent orbits in homogeneous spaces of $\SL(2,\R)\ltimes(\R^2)^{k}$

Let $G=\SL(2,\R)\ltimes(\R^2)^{k}$, let $\Gamma$ be a congruence subgroup of $\SL(2,\Z)\ltimes(\Z^2)^{k}$, and let $u_{\R}=(u_x)_{x\in\R}$ be the one-parameter subgroup of $G$ given by $u_x=\left(\matr 1x01,0\right)$. We prove polynomially effective asymptotic equidistribution results for expanding translates of $u_{\R}$-orbits and for long pieces of individual $u_{\R}$-orbits in $\Gamma\backslash G$. An important ingredient of the proof is the delta symbol version of the circle method.
Read the full abstract

Probability

Exact solution of the ruin problem in the Cramér–Lundberg with proportional investments and the limits of applicability of the power-law asymptotics

The Cramér-Lundberg model with exponential claims and proportional investment is solved exactly: the integro-differential equation for the survival probability reduces to a doubly confluent Heun equation, yielding an explicit solution in terms of Heun functions, a verification theorem, and a qualitative analysis of ruin probability versus investment share.
Read the full abstract

Optimization and Control

Deterministic Adam-Inspired Methods with Accelerated Convergence Rate

Adam is widely used, but its convergence theory remains incomplete even in the deterministic full-batch setting because momentum and adaptive preconditioning are tightly coupled. For smooth convex objectives, we split the momentum variable through variable-and-operator splitting, which reveals the acceleration mechanism. We then combine a Hessian-driven correction with Adam-style feedback based on the gradient magnitude. The resulting Adam-HNAG (Hessian-driven Nesterov accelerated gradient with Adam-style adaptive preconditioning) flow admits a nonnegative energy that decays exponentially. Its discretization yields two methods, Adam-HNAG and the synchronous variant Adam-HNAG-s. Under the stated trajectory-bound and consistency conditions, both methods satisfy a discrete Lyapunov contraction. If the exact adaptive steps are accepted, this contraction gives an $O(k^{-2})$ objective-value bound. Numerical experiments illustrate their behavior. These results apply to the proposed methods, not to the original Adam recursion.
Read the full abstract

Representation Theory

Jordan-Chevalley decompositions over imperfect fields

We give a classification of Jordan-Chevalley decompositions of an endomorphism of a finite-dimensional vector space over a not necessarily perfect field, i.e. additive decompositions into commuting semisimple and nilpotent endomorphisms.

Differential Geometry

A Conformally Invariant Dirac-type Equation on Compact Spin Manifolds: the Effect of the Geometry

Given a closed Riemannian Spin manifold $(M,g)$ of dimension greater or equal than four, we consider a generalized conformally invariant equation involving the Dirac operator with a non-linearity of convolution type. We show that the Aubib-type inequality corresponding to the problem is always strict, unless $(M,g)$ is conformal to the round sphere. In particular, this result provides an existence result for a ground state to the conformal Dirac-Einstein problem in dimension four. We point out that aside from some perturbative or special cases, this presents the first general existence result for the conformal Dirac-Einstein equations in dimension four.
Read the full abstract

Analysis of PDEs

Existence and uniqueness of weak solutions to singular anisotropic elliptic problems

In this paper we consider a quasilinear singular anisotropic elliptic problem with a p-sublinear perturbation. We prove uniqueness of weak solutions and we provide necessary and sufficient conditions for the existence of weak solutions in the same spirit of the celebrated paper by Lazer and McKenna.
Read the full abstract

Functional Analysis

Weak$^*$-weak points of continuity on the state spaces

Let $X$ be a Banach space. For $x \in X$ with $\|x\| = 1$, we denote the state space by $S_x = \{x^* \in X^* : \|x^*\| = x^*(x) = 1\}.$ In this paper, we study weak$^*$-weak and weak$^*$-$\|\cdot\|$ points of continuity of the identity map on the state spaces in the space $\ell^p(X)$ for $1 < p < \infty$, where $X$ is a non-reflexive Banach space. We then use these results to characterize the weak and norm compactness of the state spaces of unit vectors in $\ell^p(X)$. In addition, we address an open problem concerning the characterization of weakly compact state spaces in the space of Bochner-integrable functions $L^1(\mu, X)$. We also provide a local solution to this problem without any additional assumptions on the Banach space $X$. Motivated by the work of S. Daptari, V. Montesinos, and T. S. S. R. K. Rao, we show that if the set of all weak$^*$-weak points of continuity of $L^1(\mu, X)_1^*$ is weakly dense in $L^1(\mu, X)_1^*$, then $X^*$ has the Radon-Nikodým property (RNP).
Read the full abstract

Number Theory

Seven squares from three numbers

We study triples {a,b,c} of distinct nonzero rational numbers such that a+1,b+1,c+1,ab+1,ac+1,bc+1 and abc+1 are all perfect squares. We prove that there exist infinitely many such triples. In contrast, we show that no triple of positive integers has this property.

Group Theory

Subindices and subfactors of infinite groups and numbers

The theory of subfactors of groups, together with the associated notions of subindices and index stability for groupsandtheirsubsets, hasrecentlybeenintroducedandsystematicallydeveloped. Theseconceptsexhibitdeepconnections with additive combinatorics and number theory, relating to important topics such as packing and covering numbers, syndetic sets, group diameters, special integer sequences (e.g., primes and Fibonacci numbers), and classical rational sequences (e.g., Bernoulli numbers). Following the initial paper presented in 2020, two subsequent works further investigated these ideas within the framework of finite groups. In the present paper, in addition to advancing several aspects of the topic, we focus on infinite groups, with particular emphasis on groups of numbers. In this context, we introduce the RSFA (Right Subfactor Algorithm) for infinite groups and resolve several previously open problems. One of the important results is that every infinite group is index-unstable. We also correct several earlier inaccuracies and establish a weak version of a conjecture concerning differences of prime numbers. Furthermore, we determine the exact subindices for several notable sequences of integers and provide a general criterion for index stability and non-index stability of subsets in countable groups. Finally, we investigate the index stability of infinite groups and present a collection of related projects, problems, questions, and conjectures.
Read the full abstract

Analysis of PDEs

Proving the existence of localized patterns, periodic solutions, and branches of periodic solutions in the 1D Thomas model

In this paper, we present a general framework for constructively proving the existence of stationary localized solutions, spatially periodic solutions, and branches of spatially periodic solutions in the 1D Thomas model. Specifically, we develop the necessary analysis to compute explicit upper bounds required in a Newton–Kantorovich approach. Given an approximate solution $\bar{\mathbf{u}}$, this approach relies on establishing that a well-chosen fixed point map is contracting on a neighborhood $\bar{\mathbf{u}}$. For this matter, we construct an approximate inverse of the linearization around $\bar{\mathbf{u}}$, and establish sufficient conditions under which the contraction is achieved. This provides a framework for which computer-assisted analysis can be applied to verify the existence and local uniqueness of solutions in a vicinity of $\bar{\mathbf{u}}$, and control the linearization around $\bar{\mathbf{u}}$. Furthermore, as the Thomas model has a non-polynomial nonlinearity, we will need to use different techniques to handle it during our analysis. Our contributions are to provide a partial answer to how one can approach rigorously verifying results in the Thomas model, to adapt and combine previously developed techniques to apply to the Thomas model, and to perform the computer-assisted analysis to obtain such results. The code to perform the rigorous proofs is available on Github.
Read the full abstract

Numerical Analysis

$h$-$\gamma$ Blossoming, $h$-$\gamma$ Bernstein Bases, and $h$-$\gamma$ Bézier Curves for Translation Invariant $\left(\gamma_{1},\gamma_{2}\right)$ Spaces

A $\left(\gamma_{1}, \gamma_{2}\right)$ space of order $n$ is a space of univariate functions spanned by $\left\{\gamma_{1}^{n-k}(x), \gamma_{2}^{k}(x)\right\}_{k=0}^{n}$. A $\left(\gamma_{1}, \gamma_{2}\right)$ space is said to be translation invariant if $\gamma_{1}(x-h)$ and $\gamma_{2}(x-h)$ can be expressed as nonsingular linear combinations of $\gamma_{1}(x)$ and $\gamma_{2}(x)$. Translation invariant $\left(\gamma_{1}, \gamma_{2}\right)$ spaces include polynomials $\left(\gamma_{1}(x)=1, \gamma_{2}(x)=x\right)$, trigonometric functions $\left(\gamma_{1}(x)=\cos x, \gamma_{2}(x)=\sin x\right)$, hyperbolic functions $\left(\gamma_{1}(x)=\cosh x, \gamma_{2}(x)=\sinh x\right)$, and their discrete analogues. We merge $\gamma$-blossoming for $\left(\gamma_{1}, \gamma_{2}\right)$ spaces with $h$-blossoming for $h$-Bernstein bases and $h$-Bézier curves to construct a novel $h$-$\gamma$ blossom for translation invariant $\left(\gamma_{1}, \gamma_{2}\right)$ spaces generated by two continuous, linearly independent functions $\gamma_{1}$ and $\gamma_{2}$. Based on this $h$-$\gamma$ blossom, we define $h$-$\gamma$ Bernstein bases and $h$-$\gamma$ Bézier curves and study their properties. We derive recursive evaluation algorithms, subdivision procedures, Marsden identities, and formulas for degree elevation and interpolation for these $h$-$\gamma$ Bernstein and $h$-$\gamma$ Bézier schemes.
Read the full abstract

Statistics Theory

Planted clique detection and recovery from the hypergraph adjacency matrix

Hypergraph data are often projected onto a weighted graph by constructing an adjacency matrix whose $(i,j)$ entry counts the number of hyperedges containing both nodes $i$ and $j$. This reduction is computationally convenient, but it can lose information: distinct hypergraphs may induce the same matrix, and the matrix entries are generally dependent because each hyperedge contributes to multiple pairs. We study the planted clique problem under this matrix-only observation model. For detection, we show that a spectral norm test is asymptotically powerful at the $\sqrt{n}$ scale, with explicit dependence on the background hyperedge probability $p$. For recovery, we analyze a polynomial-time spectral method based on the leading eigenvector and prove exact recovery at the canonical $\sqrt{n}$ scale, again with explicit dependence on $p$. We also extend both results to sparse regimes in which the hyperedge probability may depend on \(n\). Our analysis adapts a leave–one–out eigenvector framework to this setting. These results provide rigorous detection and recovery guarantees when only the adjacency matrix is observed.
Read the full abstract

2604.08691 · 9 Apr · 45 pages. This revision fixes a measurability issue in the leave–one–out proof by separating a measurable eigenvector representative from the subsequent sign choice. It also removes an unnecessary factor left over from an earlier modification, which makes the argument more transparent

Analysis of PDEs

Instantaneous blowup and non-uniqueness of smooth solutions of MHD

We construct a family of solutions $(u,B)$ of the incompressible magnetohydrodynamic (MHD) system, the $L^\infty$ norm of which blows up instantaneously at the critical rate. The solutions remain smooth except at the blowup time. An inverse energy cascade mechanism and a convex integration scheme along a time sequence are the main ingredients of the construction, inspired by our recent work [CDP25] for the Navier-Stokes equations. The challenge of the construction for the MHD system stems from the coupling and the necessity of preserving the same ansatz of the principal solution at every iterative step while implementing convex integration. Existing convex integration schemes for MHD can treat the coupling but fail to produce the same ansatz of the principal solution recursively. To achieve the goal, we introduce a coupled geometric lemma that decomposes a symmetric tensor and a skew-symmetric tensor simultaneously. We emphasize that such coupled geometric lemma is new and of independent interest.
Read the full abstract

Optimization and Control

Stability for the stochastic heat equation with multiplicative noise via finite-dimensional feedback

In this paper, we study the long-time behavior of a stochastic heat equation with a class of infinite-dimensional multiplicative noises and localized control. We first illustrate through simple examples of the uncontrolled dynamics that mean-square and almost sure exponential stability may hold under different conditions on the parameters, reflecting the interplay between the drift and the multiplicative noise. We then construct a finite-dimensional feedback control acting on a measurable subset of positive measure, built from finitely many Fourier modes of the solution. In particular, we show that the number of controlled modes determines the decay rate and allows for arbitrarily fast stabilization in the mean-square sense. As a consequence, almost sure exponential stability is recovered via a probabilistic argument, so that both notions of stability are achieved within the same framework and with the same guaranteed decay rate. We also establish null-controllability at any positive time through an iterative construction of adapted controls from finite-dimensional feedback laws. The construction relies on the quantitative stabilization estimates and requires no observability inequality for the adjoint equation.
Read the full abstract

Rings and Algebras

A necessary condition for solvability by radicals

This note was prepared as a handout for the MAT401 course “Polynomial equations and fields", taught at the University of Toronto in Spring 2026. It presents a proof of a necessary condition for the solvability of algebraic equations by radicals, based on Galois theory. We begin with a brief overview of the relevant basic results from Galois theory, as covered in MAT401, and use – without proof – several standard (and relatively simple) results from the course textbook [1]. The sufficient condition for solvability by radicals, which is based on linear algebra, we will present in the next handout.
Read the full abstract

Number Theory

Galois theory for finite fields

This note presents Galois theory for finite fields. It was written as a handout for the MAT401 course “Polynomial equations and fields” taught at the University of Toronto in Spring 2026. We use without proofs some basic properties of finite fields and of finite field extensions which we already covered in class. Firstly, we describe an extension $K\subset F$ of a finite field $K$ of a given degree $n$. We show that the set of all intermediate fields for this extension is in one-to-one correspondence with the set of all divisors $k$ of the degree $n$. Then we describe the Galois group of this extension which is the cyclic group of order $n$. The set of subgroups of this group also is in one-to-one correspondence with the set of all divisors $k$ of the degree $n$. It allows us to prove the Galois correspondence for that extension. In the last section, we state basic theorems of Galois theory for arbitrary fields which will be proven later in the course.
Read the full abstract

Logic

Ordered henselian valued fields: definability and Borel sets

We firstly show that due to their resplendency ordered henselian valued fields admit relative field quantifier elimination in the Denef–Pas language expanded by linear orders in the field and residue field sort. Secondly, we deduce from a dimensionality reduction theorem that any set definable over an ordered henselian valued field is a Borel set with respect to the order topology. Our results are contextualised within Shelah's classification conjecture of NIP fields and its connections to the study of definable henselian valuations and the Fundamental Theorem of Statistical Learning.
Read the full abstract

Information Theory · Cross-listed from cs.CR

Realisation-Level Privacy Filtering

We study differentially private data release, where a database is accessed through successive, possibly adaptive queries and mechanisms. Existing composition theorems and privacy filters almost always combine worst case per-round privacy parameters, leaving room for more refined accounting based on realised leakage, which we term realisation-level accounting. We present a realisation-level filtering approach to determine stopping times for data releases, and design one such filter. Despite technical challenges arising from conditioning on realisations and stopping time, we prove that the filter guarantees $(\epsilon, \delta)$-differential privacy, with $\epsilon$ and $\delta$ chosen by the data handler. Through numerical evidence, we demonstrate that realisation-level filtering provides a path to better utility beyond mechanism-level methods. Furthermore, our proposed filter applies to arbitrary mechanisms, including those that are badly behaved under Rényi differential privacy.
Read the full abstract

Statistics Theory · Cross-listed from stat.ML

A Transport Geometry of Benign Overfitting

Benign overfitting, the phenomenon that interpolating noisy data can be statistically harmless, is usually characterized through effective-rank conditions on the covariance spectrum. These conditions are fragile: a perturbation of vanishing relative operator and Frobenius norm can turn a benign covariance into one whose minimum-norm interpolator has non-vanishing excess risk. We argue that the natural geometry of benign overfitting is optimal transport of the log-eigenvalues, measured in units of the sample size $n$. In log-spectral coordinates, the deterministic equivalent of the risk becomes a functional of two measures smoothed by a logistic kernel centred at the interpolation threshold $\log\kappa$. We compute its exact first variation: the threshold moves by a sensitivity-weighted average of the log-eigenvalue displacements, and the influence of each eigenvalue decays exponentially with its log-distance to the threshold. We then derive global, dimension-free stability bounds. The variance functional is $8/(27n)$-Lipschitz in the 1-Wasserstein distance between log-spectra, with a sharp constant, and is controlled more generally by a threshold-localized transport discrepancy; an analogous bound controls the bias. Consequently, any rank-$k$ update of the covariance, of arbitrary norm, changes the variance functional by at most $3k/n$, and benign overfitting survives every perturbation whose localized transport cost is $o(n)$. The results extend to ridge regression and to infinite-dimensional trace-class covariances. Matching lower bounds show that the $n$-scaling is necessary and cannot be replaced by operator norms, Frobenius norms, or dimension-normalized spectral distances. Simulations, one-step feature learning in two-layer networks, and random features of real images support the theory and delineate its limits.
Read the full abstract

Probability

Central limit theorem for the Allen-Cahn equation with supercritical random initial conditions

We study the large-scale behavior of solutions to the Allen-Cahn reaction-diffusion equation with Gaussian initial data. We consider the case of short-range dependence in the associated supercritical regime with spatial dimension $d \ge 3$. Under diffusive rescaling, the non-linearity formally vanishes on large scales in this case. Accordingly, we prove a central limit theorem for the rescaled solution, more precisely, that it converges to the solution of the heat equation started from a white noise. These initial conditions for the limit depend non-trivially both on the source of randomness and on the non-linearity. Our proof uses estimates obtained by a combination of comparison principles and Malliavin calculus, initiated by Castillo and Dunlap in arXiv:2509.06260 in the critical case. However, the result there is not a fluctuation result but rather an $L^2_\mathbf{P}$ comparison to a McKean-Vlasov problem with Gaussian solutions. Hence the mechanism behind the Gaussianity of the limit differs, and the proof requires new ideas that should be further applicable to other supercritical problems.
Read the full abstract

Analysis of PDEs · Cross-listed from physics.flu-dyn

Inverse Energy Cascade in Turbulent Taylor-Couette Flows

The inverse energy cascade in turbulent Taylor-Couette flow is studied in line with the results of the large eddy simulation. The simulation results show that the inverse energy cascade first occurs within the core region of the flow channel of the Taylor-Couette flow at higher Reynolds number. It is uncovered that this phenomenon is induced by the pulsed zero shear stress resulting from the singularities of the Navier-Stokes equation. In the core area between the two cylinders, the shear stress is nearly zero at higher Reynolds number. The turbulence generated there has high turbulent energy due to discontinuity of the tangential velocity. Since the energy transfer between the fluid layers is inhibited due to the low shear stress, the turbulent energy cannot be transferred along the radial direction, and small-scale vortices with high turbulent energy are produced. These small-scale vortices are located with the large-scale vortices and cannot be dissipated owing to low shear stress. A peak in the energy spectrum at middle frequency (or wave number) is formed due to the concentration of the small-scale vortices. As the number of the singular points of the Navier-Stokes equation increases with the increasing Reynolds number, the region with zero shear stress expands along the radial direction, intensifying nonlinear instability and energy accumulation. This, in turn, leads to more prominent peaks in the energy spectrum, resulting in a more pronounced inverse energy cascade.
Read the full abstract

All authors Changquan Zhou, Hua-Shu Dou, Lin Niu, Wenqian Xu