Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Quantum Physics

Attainable and unattainable quantum precision limits in ultrastrong light–matter coupling

In this paper, we study the gauge dependence of quantum precision in ultrastrongly coupled light–matter systems. The unitary between gauges carries the coupling $g$, so each truncation defines a different family. The quantum Fisher information traces an exact parabola across frames. Harmonic matter minimizes it at a universal $\theta_\ast=\Omega/(\Omega+\omega)$, set by the matter and cavity frequencies. Only the minimal-coupling bound is attainable. Frames differ by a factor $9.6$ with opposite trends. Diverging virtual-excitation gain at the Dicke point is a gauge artifact and saturates.
Read the full abstract

Astrophysics of Galaxies

The Connection between Halo Assembly History and the Stellar Distribution II: A Radially Stratified Fossil Record of Past Halo Mass

The spatial stellar distribution in massive central galaxies preserves a stratified fossil record of their host halos' assembly history. Using a halo-mass-complete sample from the IllustrisTNG-300 hydrodynamical simulation at $z=0.4$, we quantify this connection with two diagnostics. The fossil correlation $\tilde{\rho}(z)$ measures how strongly a present-day galaxy property retains information about the progenitor halo mass at each redshift. The rank retention fraction $f_{\rm ret}(z)$ measures how many halos selected by that property today were among the most massive halos at each epoch. The central stellar mass at different galactocentric radii exhibits non-monotonic fossil correlations with past halo mass, peaking at intermediate redshifts that systematically shift from $z\sim1$ in the inner regions to $z\sim0.7$ in the outer stellar halos. These peak redshifts are robust from group to cluster scales. Partial correlations with halo accretion rate show that each peak marks the end of the epoch when that component's buildup was sensitive to halo accretion. The rank retention fraction shows that the annular stellar masses of the central galaxy beyond 10 kpc are less biased tracers of the most massive halos at intermediate redshifts than richness. This stratified fossil record originates mainly from the ex-situ phase of galaxy assembly. The outer stellar halo also traces the satellite population at intermediate redshift better than present-day richness. These results establish that the stellar profile of massive central galaxies encodes a radially stratified fossil record of halo assembly history.
Read the full abstract

All authors Shuo Xu, Song Huang, Conghao Zhou, Alexie Leauthaud, Benedikt Diemer

Rings and Algebras

Dualities and completions of median algebras and semilattices: a universal-algebraic approach

We show that universal algebra, and the theory of natural dualities in particular, provides a uniform framework for the dualities between median algebras and bounded poc sets and their Boolean topological variants. We identify the Roller compactification of a CAT(0) cube complex with the profinite completion and the natural extension of the median algebra of its vertices. We then develop the order-theoretic perspective, which leads to a characterization of the profinite completion by the property that its intervals are the canonical extensions of the original ones. Since the basepoint orders of a median algebra have no top in general, we introduce conditional canonical extensions, show that they are unique, and realize them inside the profinite completion. Finally, a transfer theorem carries universal sentences from the finite quotients of a median algebra to its profinite completion, and we close with open questions on a correspondence theory between median algebras and their bounded poc sets.
Read the full abstract

Computational Geometry

Sets of reach $R$ are $\operatorname{CAT}\left( 1/R^2\right)$

Positive reach underpins many results in computational geometry and topology. It is used for triangulation criteria, topological inference, and manifold learning. The geometric properties of these sets have therefore been studied intensely. In this paper we focus on the intrinsic metric properties of sets of positive reach. In particular we give an elementary proof of the fact that (closed) sets of reach at least $R$ are $\operatorname{CAT}(1/R^2)$. This fact was established earlier by Alexander and Bishop, using quite significant machinery. The fact that a space is $\operatorname{CAT}(1/R^2)$ offers many advantages, such as the uniqueness and continuity of sufficiently short geodesics, the existence of Fréchet means and by extension the Rips based homotopy reconstruction of Hausmann is available.
Read the full abstract

Number Theory

An asymptotic formula for primes of the form $u^2+v^2+1$

We prove Motohashi's conjecture on primes of the form $u^2+v^2+1$, with Iwaniec's corrected constant. Let $b$ denote the characteristic function of the positive integers representable as a sum of two squares. Then, as $x\to\infty$, \[ \sum_{p\le x}b(p-1) =\kappa\Spr\frac{x}{(\log x)^{3/2}} +O\!\left(\frac{x}{(\log x)^{3/2}\sqrt{\log\log x}}\right), \] where $\kappa$ is the Landau–Ramanujan constant and $\Spr=\prod_{p\equiv3\,({\rm mod}\,4)}\bigl(1-1/(p(p-1))\bigr)$. The proof combines auxiliary prime factors with a mean square estimate for bilinear sums. We obtain the minor arc bound through harmonic switching and high moments, replacing arithmetic divisibility conditions by independent projective lines. We remove the auxiliary weights using Selberg's sieve.
Read the full abstract

Machine Learning (Statistics)

Prediction with Foundation Models under Deployment Budgets

Foundation models can supply predictions for a downstream task that draw on knowledge from extensive pretraining, but they differ in task-specific accuracy and query cost. We develop a framework that improves prediction for the task under a budget on the average query cost per input, using training data to learn which models to query and how to combine their outputs with features available without querying. The first method queries the same models for every input; the second lets them vary across inputs and uses Q-learning to decide which model to query next and when to stop. For the first method, we show that several individually inaccurate outputs can jointly outperform the best single one. It attains the minimax rate when the regression function depends on the outputs only through models within the budget. Otherwise, any method within the budget can be forced to a slower rate, which the first method attains. For the second method, we show when varying the queried models across inputs strictly improves on querying the same models for every input, and bound its prediction risk and average cost in finite samples. In simulations and a multimodal laptop price application, both methods outperform existing methods.
Read the full abstract

Statistical Mechanics

Giant Diffusion at Flocking Transitions

We simulate dynamics in two-dimensional aligning active-matter models exhibiting order-disorder transitions in the velocity field. Analysis reveals that the single-particle diffusivity, $D_s$, exhibits strong enhancements at criticality which we quantify via a finite-size scaling method. This dynamical behavior at nonequilibrium flocking transitions is fundamentally different from that at the criticality of equilibrium systems, where diffusivities either vanish or remain non-critical. Furthermore, $D_s$ exhibits pronounced finite-size effects throughout the ordered phase, which also are absent in equilibrium settings. We interpret these counterintuitive transport anomalies within the framework of a flocking theory. Our results suggest that enhanced diffusion generates effective long-range interactions. This relaxes the constraints of the Mermin-Wagner theorem that otherwise prevents order-disorder transitions in two dimensions for systems possessing continuous symmetry. These results are of much experimental relevance and have important implications in formulating quantitative theories for nonequilibrium transports.
Read the full abstract

Computation and Language

Phrase-Level Speculative Decoding on a Frozen Target

Speculative decoders raise throughput by banking more target tokens per forward pass, and Medusa, EAGLE-3, DFlash and DSpark all do it the same way: more candidates that survive verification. All of them draft in the target's own vocabulary, so one node of the candidate tree is one target token, and the tokenizer makes that unit look like a constant. It is not. BPE's pretokenizer splits on whitespace before merges are learned, so no multi-word string can ever become a single token: in Qwen3-4B, "United States of America" is four tokens and no adjacent pair of them exists as one entry, so a phrase is out of reach of any method drafting in the target's vocabulary. We change the unit. A small MLP head reads the frozen target's hidden state and predicts a one- to three-word phrase as one entry in a separate draft vocabulary; greedy longest-match expansion maps it back into the target's own tokens, and N candidates are verified in one forward pass behind a prefix-tree mask, so the change is orthogonal to tree depth and width. The target and its tokenizer stay frozen and a candidate is accepted only while it matches the target's own argmax, so the served output is identical to greedy decoding up to floating-point tie-breaking. On one A5000 at batch 1, against HuggingFace generate(), a 21.94M head (0.545% of Qwen3-4B) reaches 2.391x on GSM8K, 2.444x on MATH-500 and 1.544x on MT-Bench. Because the head's output layer is one row of 257 parameters per entry, the draft vocabulary size is the parameter budget: an 8,000-entry head is 3.434M parameters, 0.0854% of the target and 64x smaller than EAGLE-3's drafter, and keeps 92.3% of the larger head's GSM8K tokens per pass. With EAGLE-3 run on same prompts and baseline, our 21.94M head is ahead of tree-16 on both maths sets and behind tree-60 on every set.
Read the full abstract

Analysis of PDEs

An almost sharp $L^{2}(\mathbb{R}^{d})$-$L^{2}(\mathbb{S}^{d-1})$ restriction theorem for a class of anisotropic operators

Let $d\geq 2$. For each $j\in \{1,\dots,d\}$, let ${\ell}_{j}\geq 1$ and $m_{j}\geq 2$ be integers, and let $a_{j}$ be a nonzero real number. We prove that the oscillatory integral operator \begin{equation*} T_{{\ell}_{1},\dots,{\ell}_{d},m_{1},\dots,m_{d}}f(x):=\int_{\mathbb{R}^{d}} e^{\dot{\imath} \left(a_{1}x^{{\ell}_{1}}_{1}y^{m_{1}}_{1}+\dots +a_{d}x^{{\ell}_{d}}_{d}y^{m_{d}}_{d} \right)}f(y)dy,\;\;\; x\in \mathbb{R}^{d}, \end{equation*} satisfies the restriction estimate \begin{equation*} \|T_{{\ell}_{1},\dots,{\ell}_{d},m_{1},\dots,m_{d}} f\|_{L^{2}(\mathbb{S}^{d-1})}\lesssim \|f\|_{L^{2}(\mathbb{R}^{d})}, \end{equation*} whenever \begin{equation*} \mathcal{U}:= \min{\left\{ \min_{1\leq j\leq d}{\frac{m_{j}}{{\ell}_{j}}},\,\min_{1\leq k\leq d}{\frac{d-1}{\sum_{j\neq k} \frac{\ell_{j}}{m_{j}}+\frac{2}{m_{k}}}}\right\}}>1, \end{equation*} and we necessarily have $\mathcal{U}\geq 1$.
Read the full abstract

Numerical Analysis

VEM-bubble discretisation and iterative coupling algorithm for a fluid induced seismicity model with rate-and-state friction

This work addresses the discretisation and iterative coupling algorithm for an induced seismicity model accounting for fault reactivation and slip driven by subsurface fluid injection. The model couples single-phase Darcy flow in both the porous matrix and the fault network with a quasi-static poroelastic mechanical response in the bulk and fault slip governed by a rate-and-state friction law with radiation damping stabilisation leading to a quas-dynamic formulation. The contact mechanics problem is discretised using the Virtual Element Method with bubble stabilisation of the facewise constant Lagrange multiplier. It is combined with a Hybrid Finite Volume scheme for the mixed-dimensional Darcy flow model. The time discretisation is implicit and the nonlinear system is solved at each time step using an iterative coupling algorithm which iterates between the Darcy flow, the frictional contact-mechanics, and the state differential equation. The contraction of this iterative coupling algorithm is established with a smallness assumption on the time step independent on the mesh size. This proves convergence of the fixed-point as well as well-posedness of the system. The frictional contact-mechanical model is solved using a semi-smooth Newton algorithm based on a local nonlinear inversion of the friction rate-and-state law and a block Gauss–Seidel preconditioner exploiting the structure of the VEM-bubble-$\mathbb{P}^0$ mixed discretisation.
Read the full abstract

All authors El Hadi Zerga, Roland Masson, Jean-Paul Ampuero, Pierre Romanet

Differential Geometry

Branch Obstructions for Area-Minimizing Currents and Flat Critical Points of Weakly-Conformal Maps

We study two problems motivated by the local geometry of conformal $\phi$-curves. First, let $T$ be a $k$-dimensional locally area-minimizing integral current in Euclidean space, with $k\geq3$, codimension at least two, no boundary, and a locally conformally flat induced metric on $\operatorname{reg} T$. We prove the following branch obstruction: at any rapid-decay branch point whose planar frequency differs from $2$, every nonzero cyclic component of every cylindrical normalized tangent function has repetition multiplicity at least two. Consequently, the nonintegral-frequency points of densities $2$ and $3$ in the rapid-decay branch set have Hausdorff dimension at most $k-3$. On the full branch set, they have zero $\mathcal H^{k-2}$ measure. These conclusions apply to Euclidean image currents of conformal $\phi$-curves with locally conformally flat domains, provided the image currents satisfy the stated hypotheses. Second, for $C^j$ weakly-conformal maps between smooth Riemannian manifolds, with domain dimension $k\geq3$ and $j\geq2$, we prove that all derivatives of orders $1$ through $j$ vanish at every critical point. As a consequence, for smooth weakly-conformal maps with connected domain, we establish a curvature condition under which the map is constant or has no critical points. The flatness and continuation results apply in particular to conformal $\phi$-curves with the corresponding regularity and curvature assumptions. Finally, we examine the Borůvka cone, which illustrates the role of regularity and the distinction between critical points and branch points.
Read the full abstract

Materials Science

Revealing laser-induced modification of weak ferromagnetism through a spin-reorientation phase transition in iron borate Fe$_3$BO$_6$

In the iron oxide Fe$_3$BO$_6$ weak ferromagnetism stems from competing contributions from two-site Dzyaloshinskii-Moriya interactions (DMI) and from on-site anisotropies at the two $4c$ and $8d$ types of nonequivalent Fe$^{3+}$ sites. As a result, spin reorientation (SR) transition $\Gamma_2$ $\rightarrow$ $\Gamma_4$ in Fe$_3$BO$_6$ uniquely comprises 90$^\circ$ spin switching and a 35% change of weak ferromagnetic moment. We explore ultrafast kinetics of such a transition as a proxy for weak ferromagnetism contributions modified by femtosecond laser pulses. Using magneto-optical pump-probe technique we reveal that the SR transition is triggered by an abrupt decrease of the weak ferromagnetic moment in the $\Gamma_2$ phase, and relate it to the laser-induced modification of the DMI and anisotropy contributions to the spin canting. By tuning the pump pulse phonon energy between different optical absorption ranges, we show that laser-driven impact on spin canting possesses a nonthermal character and is achieved through the charge-transfer transitions $2p\rightarrow3d$ within Fe$^{3+}$O$^{2-}_6$ octahedra rather than by localized $3d\rightarrow3d$ transitions at Fe$^{3+}$. The decreased weak ferromagnetic moment controls the threshold and the latency of the SR transition, as well as a final state. The latter is achieved via the precessional switching to the $\Gamma_4$ phase activated by the laser-induced heating. Final state is found to be a coexistence of $\Gamma_4$ and $\Gamma_2$ phase with modified spin canting persisting even at hundreds of picoseconds.
Read the full abstract

All authors A. V. Kuzikova, S. N. Barilo, R. A. Lanovsky, A. V. Sadovnikov, R. V. Pisarev, A. M. Kalashnikova

General Economics

Fiscal Commitment and the Additionality of Philanthropy: Heterogeneous Donor Purposes in a Shared Public Budget

Philanthropic donors often value different public services, while a budget authority allocates public spending across a shared portfolio. Policy frameworks frequently seek to ensure that private finance supplements rather than displaces public funding. We distinguish baseline additionality, which preserves pre-gift public funding, from portfolio preservation, which preserves a service's provision after all gifts are pooled under discretionary budgeting. A larger gift to one service raises the discretionary provision of every service and reduces the protection slack available to the others. This cross-donor fiscal externality yields an exact feasibility result: if a service's maximum permissible floor lies below its discretionary post-gift provision, every effective protection reform reduces that service relative to the portfolio reference, including full protection of its own appropriation. Conversely, a reform that expands at least one donor-backed service while preserving all others exists exactly when every maximum floor reaches the reference. The paper characterizes the complete preservation set and its welfare-maximizing element. Under logarithmic preferences, donation shares determine protectability, while heterogeneous donor valuations determine welfare incidence. Baseline additionality may therefore hold while a service loses fiscal benefits generated by other gifts.
Read the full abstract

Machine Learning (Statistics)

Causal discovery from Bayesian symbolic regression

Causal discovery from observational data is central to scientific reasoning, yet established constraint-based and score-based methods often recover only a Markov equivalence class and provide no signal in bivariate or strongly autocorrelated regimes typical in some areas of science. Here, we recast causal discovery as a Bayesian symbolic regression problem. Given a set of candidate parents for each target variable, the Bayesian symbolic regression samples closed-form expressions from a posterior over symbolic models. The fraction of sampled expressions in which each candidate variable appears defines a posterior inclusion probability that we use as the causal score. This replaces pointwise model selection by an ensemble of plausible mechanisms. We evaluate the method on coupled nonlinear bivariate dynamics, on the CauseMe benchmark suite spanning linear, chaotic, nonlinear, higher-dimensional and high-noise settings, and on a four-variable biogeoscience structural causal model under Gaussian, heavy-tailed and mixed-noise regimes. Our approach is competitive with, and often outperforms, specialized causal discovery algorithms. Additionally, the method returns closed-form mechanisms leading to explicit causality and interpretability.
Read the full abstract

All authors Roger Guimerà, Kai-Hendrik Cohrs, Sergio Pedreira, Marta Sales-Pardo, Gustau Camps-Valls

Signal Processing

Grid Design for Near-Field Channel Estimation with Uniform Linear Arrays

This paper develops deterministic grids for line-of-sight near-field channel estimation with uniform linear arrays (ULAs) and hybrid receivers. A normalized channel approximation squared error (NCASE) criterion leads to geometric covering problems under a constant-amplitude parabolic model and an elliptical approximation of the correlation regions. For a horizontal ULA, transformation and scaling yield a disk-covering problem, addressed by adapting a hexagonal lattice to the mapped ground-plane region. For a vertical ULA, a reciprocal-distance interval cover is solved optimally by a farthest-reaching greedy procedure. Numerical evaluation with exact spherical-wave channels yields empirical maximum approximation errors close to the target and improved P-SOMP estimation relative to the tested polar-domain grids of comparable size.
Read the full abstract

Functional Analysis

The symmetric Gaussian isoperimetric problem in the plane

We resolve the symmetric Gaussian isoperimetric conjecture in the plane: every minimizer of Gaussian perimeter among centrally symmetric sets of prescribed Gaussian measure is a centered ball, a symmetric strip, or the complement of one of these sets. A quadrant-reflection argument reduces the problem to unconditional sets. The planar structural results of Barchiesi, Cagnetti and Julin then leave only a noncircular strictly convex stationary oval. We exclude this oval by an explicit volume-preserving and perimeter-decreasing linear deformation, which geometrically corresponds to an "unconditional squeezing". More generally, without requiring any symmetry assumptions, we show that any noncircular closed embedded $\lambda$-self-shrinker is always unstable via linear deformations. We also record a reduction to the unconditional case in dimension three, using an orthogonal equipartition theorem of Musin.
Read the full abstract

Analysis of PDEs

A capillary logarithmic Gauss curvature flow and the capillary Minkowski problem

We study a logarithmic Gauss curvature flow of strictly convex capillary hypersurfaces in the Euclidean half-space. For an acute contact angle and an arbitrary positive smooth prescribed function, we prove that a suitable initial dilation yields a global solution which converges smoothly after a horizontal translation with a uniquely determined constant velocity. As an application, we provide a logarithmic flow proof of the capillary Minkowski problem, previously established by Mei-Wang-Weng [Adv. Math. 469, 110230 (2025)].
Read the full abstract

Superconductivity

Real Topology on Symmetry-Invariant Loops Enforces Superconducting Pairing Zeros

Normal-state topology can obstruct superconducting pairing and give rise to gapless nodes in the quasiparticle spectrum. However, existing pairing-obstruction mechanisms largely rely on Fermi-surface topology or global Bloch-band topology. Here, we establish a new and general mechanism: one-dimensional (1D) real topology, defined on antiunitary-symmetry-invariant loops, enforces pairing zeros without requiring Fermi-surface or global Bloch-band topology. We derive sufficient conditions and lower bounds for pairing zeros across diverse pairing channels, for both zero and finite Cooper-pair momenta. The theorems apply to systems of arbitrary spatial dimensions without requiring the pairing bands to be globally isolated. We further show that such topologically enforced pairing zeros, if isolated, necessarily carry odd vorticity and can give rise to Bogoliubov-de Gennes nodes and zero-energy boundary modes, as demonstrated in a minimal two-orbital model. Our results reshape the understanding of how normal-state topology can obstruct the superconducting order, highlight the significant role of 1D real topology beyond single-particle band phenomena, and provide a general framework for identifying candidate materials hosting nodal superconductivity.
Read the full abstract

Combinatorics

Exact local spectral thresholds for loose and tight Hamilton cycles in $3$-graphs

For an $n$-vertex $3$-uniform hypergraph $H$, let $\sigma(H)$ denote the minimum adjacency spectral radius of its vertex links. We determine the exact thresholds for $\sigma(H)$ that force loose and tight Hamilton cycles for all sufficiently large admissible orders, and give constructions showing that each bound is best possible. The loose threshold is $(1+\sqrt{13})n/8+O(1)$, with two forms according to $n$ modulo four. The tight threshold is $2n/3+O(1)$, with four forms determined by $n$ modulo three and, when $3\mid n$, modulo nine. The resulting spectral conditions and the corresponding minimum vertex degree conditions do not imply one another. For loose cycles, we combine fractional packing stability and absorption with exact completion near a space barrier. For tight cycles, spectral stability and rooted sampling reduce the problem to three extremal configurations. We resolve these configurations by controlling the spectral loss caused by missing triples and constructing short paths that extend to Hamilton cycles, including in the presence of exceptional vertices.
Read the full abstract

Robotics

V3DSAM: Semantic-Geometric Loop Verification and Geometry-Anchored Persistent Instance Mapping for Feed-Forward Dense SLAM

To address unreliable cross-view association in long-sequence feed-forward dense SLAM, this letter presents V3DSAM, an open-vocabulary semantic dense SLAM system that establishes physical correspondences at both the scene and object levels. The system combines feed-forward geometric estimates from VGGT-SLAM 2.0 with open-vocabulary semantic observations from SAM3. We introduce semantic-geometric dual-scale loop verification, which identifies false loop closures using pixel-level bidirectional reprojection and object-level 3D spatial correspondences, without accessing backbone-internal features. We further propose geometry-anchored persistent instance mapping, which defines physical instances through class-constrained 3D connectivity in an optimized global coordinate system and assigns global IDs without inter-frame identity propagation. The proposed verifier rejects all four false loop candidates in our TUM RGB-D evaluation, while long-sequence ScanNet experiments show improved instance identity consistency. Thus, V3DSAM reduces dependence on internal backbone features and temporal ID propagation for robust long-term cross-view association.
Read the full abstract

All authors Maolin Ye, Dongqing Zhao, Linyang Li, Shuai Xue, Yuan Liu, Yiyuan Wang, Wen Li

Combinatorics

A proof of the Theta conjecture

We prove the touching refinement of the Theta conjecture of Interdonato and Iraci, giving a formula for $\Theta_{e_\ell}\Theta_{e_k}\nabla E_{d,j}$ as a weighted sum over parking functions with decorated rises and valleys and touch number $j$. Summing over $j$ gives the Theta conjecture. As corollaries, we prove the valley Delta conjecture of Haglund, Remmel, and Wilson and its generalized version proposed by Qiu and Wilson. Our proof compares algebraic and combinatorial recurrences using an extension of the plethystic substitution $X\mapsto(q-1)X$ to quasisymmetric functions. The algebraic recurrence follows from an endomorphism constructed by D'Adderio, Interdonato, Iraci and Pagaria on the Dyck path algebra of Carlsson and Mellit. The combinatorial recurrence is obtained by introducing signed labels and deleting the main diagonal. The two recurrences have the same initial conditions, which identifies the corresponding generating functions.
Read the full abstract

All authors Haojun Bai, Feng Gu, Zhuowei Lin, Dun Qiu, Minhao Zhang

Algebraic Geometry

$\lambda_g$-Integrals on triple, one-part admissible covers

We effectively compute all intersection numbers of the form $\lambda_g\psi^d$ on moduli spaces of genus $g$ admissible covers with one point of full ramification and two additional branch points with arbitrary profile. We encode these invariants in appropriate generating functions, which we relate to the Hamiltonians of the KdV hierarchy.
Read the full abstract

Complex Variables

Unique continuation property for $\bar\partial$ with $L^{2n}$ potentials

We establish the sharp unique continuation property for the Cauchy–Riemann operator $\bar{\partial}$ in higher dimensions. Let $0\in\Omega \subset \C^n$ be a domain. Suppose $u \in W_{\mathrm{loc}}^{1,2}(\Omega; \C^N)$ satisfies the differential inequality \[ |\bar{\partial} u| \leq V |u| \quad \text{a.e.~on } \Omega \] with $V \in L_{\mathrm{loc}}^{2n}(\Omega)$. If $u$ is $L^2$ flat at the origin, then $u \equiv 0$ on $\Omega$.
Read the full abstract

Astrophysics of Galaxies

Revisiting the Einstein Cross with X-shooter spectra: New estimates of accretion disk parameters and broad emission line profiles from microlensing

We present a spectroscopic study of the lensed quasar Q2237+0305, the Einstein Cross, assessing the best wavelength coverage to date with the X-shooter instrument (from $3000$ to $20000$ $\mathring{A}$), and thus improving the available spectroscopic data for this system. Using as a reference the characteristic emission lines for each quasar macro-image, we estimate the chromatic microlensing effect between pairs of images and calculate the extinction curve, confirming no significant differences between spectra $C$ and $A$ caused by extinction in the lens galaxy. In addition, we model the accretion disk characterized by a temperature profile with logarithmic slope $1/p$ and size $r_{s}$, assuming a thin-disk model with a power-law function of the form $r_{s}$ $\propto$ $\lambda^{p}$. To do this, we calculated a set of three chromatic microlensing measurements compiling previous estimates from the literature to test the wavelength dependence. This temperature profile was also compared to the canonical thin-disk model with $p=4/3$. Finally, we estimated the likelihood between the disk model and the observed data collected from multiple epochs through mock microlensing magnification maps and determined the probability density functions of each parameter from a Bayesian analysis. As a result, our estimated values for the accretion disk size and its exponential index are $r_{s}=4.9^{+2.3}_{-2.6}$ $\sqrt{M/0.3M_{\odot}}$ (lt-day) and $p=0.77^{+0.47}_{-0.28}$, respectively, reinforcing a departure from the standard accretion disk model ($ T \propto r_s^{-3/4}$).
Read the full abstract

All authors R. Jerez-Nicurcar, A. Esteban-Gutiérrez, V. Motta, E. Mediavilla

Numerical Analysis

Total variation error bounds for Schrödinger-Föllmer samplers

Sampling from high-dimensional, multimodal target distributions plays a fundamental role in scientific computing, statistics and machine learning. Recently, the Schrödinger-Föllmer sampler (SFS) has been proposed by Huang et al. (IEEE Trans. Inform. Theory, 2025) as an efficient sampling algorithm that transports the degenerate distribution at $t=0$ to the target distribution at $t=1$, without relying on long-time ergodicity. In this paper, we establish non-asymptotic error bounds for SFS in the total variation distance by deriving regularization estimates for the associated Kolmogorov equation via the Bismut-Elworthy-Li formula. When the drift can be computed exactly, a convergence rate of order $\mathcal{O}( h|\ln h|)$ was obtained under certain smoothness conditions on the drift with a uniform time step-size $h>0$. When the drift is approximated by Monte Carlo methods, we obtain a convergence rate of order $\mathcal{O}( h|\ln h|+ 1/M)$, where $M$ is the number of samples used in the Monte Carlo approximation. This yields a sharper dependence on $M$ than the existing bound $\mathcal{O}(1/\sqrt{M})$ in the $L^2$-Wasserstein distance in the aforementioned paper, by turning the total variation error to the weak error analysis in the setting of non-smooth test functions. Numerical experiments are presented to illustrate the convergence rate with respect to the time step-size $h$ and the Monte Carlo sample size $M$ and the effectiveness of the proposed algorithm.
Read the full abstract

Neural and Evolutionary Computing

Pricing Symbolic Rules by What GNNs Leave Unlearned

Symbolic knowledge (motif counts, ring counts, label constraints) is routinely added to graph neural networks (GNNs) and justified by their limited expressivity. We replace this yardstick with an exact one. The largest log-loss reduction a rule $S$ can give a model family is $I(Y;S\mid\sigma)$ given what the family computes; every realised gain splits exactly into this term plus repaid neural debt; and for a trained network with representation $T$ the value is $I(Y;S\mid T)$, the label information training failed to extract. Rule design thus becomes rule discovery: we score candidates on the network's out-of-fold residual against calibrated permutation nulls with family-wise error control. On ZINC, discovery recovers the hidden ring term $\max(0,\ell_{\max}-6)$ of the target in three independent runs ($p_{\mathrm{adj}}=0.001$) and beats expert cycle counts; frozen and applied to five unseen datasets it is never significantly worse than the plain network or the expert rule, and on BBBP it significantly beats both.
Read the full abstract

Computer Vision

LongSCOP: Semantically Consistent Long Video Outpainting

Video out-painting is an important problem, yet existing research focuses primarily on short clips, leaving long video out-painting largely underexplored. Extending video out-painting to long videos poses substantial challenges. A natural solution is iterative out-painting, i.e., generating videos chunk by chunk, but this introduces two major issues: generation instability or even failure caused by inter-chunk inconsistencies and error accumulation, and semantic drift over long temporal horizons. In this paper, we propose LongSCOP, short for Semantically Consistent long Video Out-Painting. First, we introduce a training recipe with three components for smooth and stable long-horizon generation: self-forcing training, latent calibration, and future references. Beyond generation stability, we recognize that the rich temporal context of long videos provides valuable references for maintaining semantic consistency. We therefore develop a VLM-based agent that automatically selects the most informative reference frames from the entire video to maintain consistent subjects, scenes, and backgrounds throughout long-horizon generation. We also construct a high-quality training dataset of 62K video clips. To facilitate research on this underexplored problem, we introduce LVO-Bench, a benchmark curated by human experts to assess long-horizon stability and semantic consistency. Extensive experiments demonstrate that our method outperforms existing approaches in both generation stability and semantic coherence.
Read the full abstract

All authors Yu Tang, Zhiyuan You, Zhixiao Wang, Fanghua Yu, Qingyu Zhang, Xiang Yin, Chao Dong, Jinjin Gu

Combinatorics

Sampling Colorings on $G(n, d/n)$ via Glauber Dynamics

We study the mixing time of Glauber dynamics for sampling uniformly from the proper $q$-colorings of sparse Erdős-Rényi random graphs $G(n,d/n)$. Our main result establishes that with probability $1-o(1)$ over the draw of $G$, Glauber dynamics is rapidly mixing for every $q\geq d+d^{1/2+\varepsilon}$, given any fixed $0<\varepsilon\leq1/2$ and any sufficiently large fixed $d$. In particular, the surplus of colors above $d$ is sublinear and close to $\sqrt{d}$. The achieved color range therefore approaches the tree uniqueness threshold and improves over the previously best achievable number of colors $q > 1.763 d$ due to Efthymiou, Hayes, Stefankovic, and Vigoda [SODA 2018]. Moreover, our mixing-time bound is $O(n^{2+o_d(1)})$, and hence almost quadratic as $d\rightarrow\infty$. Our proof is based on establishing a spectral gap for block dynamics resulting from a suitable decomposition of $G(n,d/n)$ into blocks and singleton vertices, ensuring that every high-degree vertex is encapsulated within a block. Our main technical contribution is a novel way of establishing rapid mixing of block dynamics via an integrated Bochner-Bakry-Émery approach, which we expect to have applications beyond sampling colorings. We further devise a new block decomposition of $G(n,d/n)$ that strengthens and simplifies similar constructions used in previous work. Finally, to compare block and Glauber dynamics, we establish a novel recursive separator decomposition of tree-like blocks, together with a coupling argument that avoids the need for nearly uniform color marginals at vertices adjacent to hubs. This comparison argument might be of independent interest.
Read the full abstract

Computer Vision

MoSTE: Mixture of Spatio-Temporal Experts for Deepfake Detection in Videos

Generalizable deepfake video detection remains challenging because forgery artifacts vary substantially across manipulation types and often combine diverse spatial and temporal inconsistencies. Existing parameter-efficient adaptation methods typically rely on a single shared representation, rendering them prone to manipulation-specific bias and limiting transfer to unseen forgeries. To address this challenge, we propose Mixture Of Spatio-Temporal Experts (MoSTE), an adaptation framework that enables conditional specialization of forensic representations. Instead of forcing one adapter to model all forgery patterns, MoSTE decomposes adaptation into parallel spatial and temporal expert pools and dynamically routes tokens to the most relevant experts through an unsupervised token-level routing mechanism. This allows experts to naturally specialize in complementary forgery cues without requiring explicit manipulation labels or multi-stage training, while effectively transferring representations from CLIP to the video domain. Extensive experiments across cross-dataset and cross-manipulation benchmarks demonstrate that MoSTE sets a new state of the art in average generalization while preserving strong performance across diverse manipulation families. Additional analysis of expert routing and attention behavior reveals the emergence of expert diversity, suggesting that MoSTE learns complementary spatio-temporal forensic features. Code is available at https://github.com/charbelyah/MoSTE
Read the full abstract

All authors Charbel Yahchouchi, Nabyl Quignon, Noemi Roggero, Laurent Saroul, Antitza Dantcheva

Networking

When Handshakes Go Dark: An Adaptive Semantic-Sequential Fusion Framework for Encrypted Traffic Classification

Encrypted traffic classification is essential for network security monitoring and traffic management. Self-supervised pre-training can learn reusable representations from large-scale unlabeled traffic, but TLS 1.3, QUIC, and privacy mechanisms such as Encrypted Client Hello (ECH) increasingly conceal application, handshake, and service-identifying information. Observable upper-layer bytes are therefore dominated by ciphertext, weakening the link between raw-byte patterns and application semantics and limiting model generalization. Moreover, the availability and discriminative value of observable information vary across protocols, protection profiles, and tasks, complicating classification across diverse scenarios. To address these challenges, we propose FLARE, a Field-aware Learning and Adaptive Representation Fusion framework for encrypted traffic classification. FLARE pre-trains a BERT-based packet encoder through Masked Field Prediction, which reconstructs masked tokens from field-structured context. During downstream fine-tuning, the parameters of the pre-trained packet encoder remain frozen, and a packet-level gated module adaptively combines protocol-semantic representations with directional packet-length representations. A flow-level encoder then models inter-packet dependencies for classification. We evaluate FLARE on eight datasets spanning diverse encrypted-traffic tasks, including CipherSpectrum and VisQUIC for strongly encrypted scenarios. FLARE achieves superior performance over baselines across the eight datasets, exceeding the strongest pre-trained baseline by 10.92 percentage points on QUIC website fingerprinting. These results demonstrate the effectiveness of field-aware pre-training and adaptive semantic-sequential fusion under constrained observability.
Read the full abstract

All authors Tong Guan, Yunchun Li, Minyuan Xiao, Mingyuan Xia, Yuchen Zhao, Wei Li