Logic in Computer Science
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 8 November
We study the decidability of modal satisfiability over classes of Kripke frames defined by universal first-order sentences. Although some such classes yield undecidable satisfiability problems, decidability is guaranteed for wide families of frame conditions, including universal Horn sentences and universal sentences implying transitivity. A more than decade-old open problem asks a natural question: can decidable and undecidable cases be distinguished algorithmically? We prove that this is impossible, already for equality-free universal frame conditions over a single binary relation. More precisely, there is no algorithm that, given such a condition, decides whether its associated local modal satisfiability problem is decidable. We first establish the result for conditions with equality and then eliminate equality by passing to quotients of worlds with identical incoming and outgoing edges.
Analysis of PDEs
On the sharpness of a Liouville exponent in a competitive elliptic Schrödinger system
We identify $p_*(n)=(n-1)/(n-3)$, $n\geq 5$, as the sharp exponent for Liouville theorems concerning a two-component elliptic Schrödinger system with self-focusing interactions and repulsive intercomponent coupling. We consider the strictly copositive regime, in which the repulsion strength is smaller than the geometric mean of the two self-interaction strengths. Quittner and Souplet (Comm. Math. Phys. 2012: 1–19) established a Liouville-type nonexistence result in the range $1<p<p_*(n)$. For every $p_*(n)<p<(n+2)/(n-2)$, we construct bounded, nontrivial, nonnegative entire solutions whenever the repulsion strength approaches the copositivity threshold from below. The construction combines a compact variational functional, a periodic spectral gap, and uniform exterior estimates that remove a truncation. Our result gives a negative answer to the open question posed by Quittner and Souplet (Comm. Math. Phys. 2012: 1–19) of whether strict copositivity implies nonexistence of nontrivial nonnegative entire solutions throughout the Sobolev subcritical range when $n\geq 5$.
Read the full abstract
Information Retrieval
Dhara: Evidence-Oriented Retrieval for Indian Knowledge Systems
Relevant evidence in Indian Knowledge Systems (IKS) collections may be distributed across passages in different parts of a work, including verses, translations, and commentaries. Locating it requires matching natural-language questions and partially remembered verse fragments to source passages. We present Dhara, a retrieval engine combining hybrid search, chapter- and book-level Entity Notes (EN), Fuzzy Verse Retrieval (FVR), and neural reranking. Five comparisons on the 325 questions in Dhara Bench examine retrieval representations against BM25, EN's contribution, reranker quality and computation time, FVR, and Gemini. BGE-M3 single-vector retrieval exceeds BM25 in nDCG@10, Recall@20, and MRR@20 across all categories. Its scores are broadly comparable to ColBERTv2, with mean retrieval times of 0.108 s and 0.962 s under the same hardware configuration and a smaller memory footprint for chunk storage on disk. Adding EN to a direct branch with a matched candidate count improves nDCG@10 in all twelve categories with Qwen and ten with BGE; the combined ranking is capped at the same candidate count. Qwen improves nDCG@10, Precision@5, and MRR@20 over BGE in every category, at approximately 6.72 times the overall estimated serial computation time. FVR increases Verse-category nDCG@10 by approximately 20 percentage points with either reranker. Gemini is evaluated after source mapping and manual review with relevance annotations visible. The Qwen pipeline with EN and FVR has higher Recall@20 than Gemini in ten categories and stronger complete-evidence retrieval in Multi-hop questions, while Gemini retains early-relevance advantages in several categories. The results inform retrieval choices according to evidence requirements and resource budgets.
Read the full abstract
Optimization and Control
Optimal Dividends and Voluntary Liquidation under Omega-Type Regulatory Termination
We study a joint optimal dividend and voluntary liquidation problem for a company whose surplus follows a spectrally negative Lévy process and is subject to Omega-type regulatory termination. Dividend payments are modeled as a singular control, while voluntary liquidation is an optimal stopping decision with a state-dependent liquidation payoff. When the surplus falls below a regulatory solvency threshold, an Omega clock accumulates the time spent in distress and may terminate the firm before voluntary liquidation occurs. Under dynamic programming and continuity assumptions for an auxiliary Markov subproblem, we characterize its value as the smallest controlled stochastic supersolution of the associated quasi-variational inequality. We prove that any regular stochastic solution with a dividend barrier coincides with the value function over all admissible dividend and liquidation policies, and hence is unique within this class. Using $\omega$-scale functions, we construct single-band, singleton, and no-liquidation candidates, establish sufficient conditions for global optimality, and give conditions under which these candidates supply the unique regular barrier stochastic solution. Brownian and Cramér–Lundberg examples illustrate how regulatory parameters affect dividend and liquidation boundaries and generate transitions between policy regimes.
Read the full abstract
Algebraic Geometry
A Lefschetz Alternative for Real Characters on Smooth Quasi-Projective Varieties
We establish a Lefschetz alternative for real characters on smooth complex quasi-projective varieties, extending Simpson's classical theorem beyond the compact Kähler setting. More precisely, for a nonzero real character of the fundamental group of a smooth connected complex quasi-projective variety, either the character factors through a holomorphic fibration onto a Riemann orbifold with connected general fiber, or every level set of an associated primitive function on the corresponding covering is connected and its inclusion induces a surjection onto the fundamental group of the covering. Our approach combines logarithmic holomorphic $1$-forms with a geometric flow construction to overcome the difficulties arising from noncompactness. Meanwhile, we prove a quasi-projective analogue of the Napier–Ramachandran theorem, showing that an epimorphism onto $\Z $ whose kernel is not finitely generated must factor through a hyperbolic Riemann orbifold. As applications, we derive several new restrictions on the structure of quasi-projective groups.
Read the full abstract
General Mathematics
Monotonicity of Wang's recursive sequence and logarithmically complete monotonicity of the reciprocal of the inverse tangent
In this paper, we study a sequence $b_n$, posed by Fei Wang, which is defined by $b_1=\frac13$ and a convolution-type recursive relation. We derive the generating functions of $nb_n$ and of $b_n$, the latter of which is $\ln\frac{\operatorname{arctanh}x}{x}$, and represent $b_n$ as a moment integral over $[0,1]$ with a positive density. From this representation we prove that the sequence $b_n$ is positive and logarithmically convex, that the sequence $nb_n$ is strictly decreasing, and that the ratio $\frac{b_{n+1}}{b_n}$ is strictly increasing in $n\in\mathbb{N}$ with limit $1$, which answers Wang's problem. We also give a determinantal expression and closed-form formulas for $b_n$ involving Stirling numbers of the first kind and partial Bell polynomials. Finally, we prove that the function $\frac1{\arctan x}$ is logarithmically completely monotonic on $(0,\infty)$.
Read the full abstract
Probability
Infinite dimensional Gaussian limits for the giant component of inhomogeneous random graphs
We prove Gaussian limits for the type measure of the giant component in sparse inhomogeneous random graphs. Countable types yield convergence in a weighted Hilbert space under boundedness, irreducibility, supercriticality, and a uniform tail condition on the empirical type weights. For types on a finite dimensional closed Riemannian manifold, smooth kernels yield convergence in a negative Sobolev space for any Borel type distribution. We allow deterministic types and first order kernel perturbations; independent type sampling adds an independent Gaussian term. The proof combines a seeded exploration identity, Hilbert space martingale limits, tail estimates, and inversion of the branching operator outside the giant. Finite types with random parameters are covered. A heat kernel example gives explicit variances for all spectral modes.
Read the full abstract
Software Engineering
Beyond Code Translation: Agentic Android-to-HarmonyOS Migration
Migrating Android applications to HarmonyOS is an important yet challenging software engineering task. Existing large language model (LLM)-based migration approaches primarily translate code at the method or file level, overlooking project-wide dependencies and workflow coordination. Moreover, ArkTS, the primary language for HarmonyOS, is relatively low-resource, and migrated applications often lack sufficient tests for reliable verification, making end-to-end migration error-prone. We present A2HTrans, an agentic framework for end-to-end Android-to-HarmonyOS application migration. Rather than performing isolated code translation, A2HTrans decomposes applications into functionality-aware migration tasks, prioritizes them through dependency analysis, and orchestrates translation, validation, and iterative repair within a unified execution workflow. To improve migration reliability, A2HTrans employs a tool-augmented verification strategy that combines compiler-based syntax checking with LLM-based semantic validation and automated repair. We evaluate A2HTrans on 14 Android applications spanning diverse sizes and domains using Claude-Sonnet-4.6 as the underlying migration model. Compared with Claude Code, A2HTrans preserves 8.3% more functional completeness, improves behavioral correctness by 8.0%, and reduces migration time by 28.8%. These results demonstrate that structured agentic execution substantially improves the reliability and efficiency of project-level cross-platform software migration.
Read the full abstract
Combinatorics
Token graphs of stars
The $k$-token graph $F_k(G)$ of a graph $G$ has as vertices the $k$-subsets of $V(G)$, two of them being adjacent when their symmetric difference is an edge of $G$. Let $S_n$ be the star with $n$ leaves. Since every edge of $S_n$ is incident with the center, $F_k(S_n)$ is the inclusion graph of the $(k-1)$- and the $k$-subsets of the leaves. We use this identification to determine its diameter, radius, girth, Wiener index and number of spanning trees, and to decompose $F_k(S_{n+1})$ into $F_k(S_n)$ and $F_{k-1}(S_n)$ joined by a matching; the parameters of $F_k(S_n)$ that are already known, such as its connectivity, independence number, automorphism group and spectrum, are collected in an appendix. The two parameters that the identification does not settle are the domination and the packing numbers. For the domination number we prove, for fixed $k \ge 3$, that $(1-\frac{k-2}{k-1}\pi_k-o(1))\binom{n}{k-1} \le \gamma(F_k(S_n)) \le (1-\frac{1}{2}(\frac{k-2}{k-1})^{k-2}+o(1))\binom{n}{k-1}$, where $\pi_k$ is the Turan density of the complete $(k-1)$-uniform hypergraph on $k$ vertices; for $k=3$ both constants are equal to $\frac{3}{4}$, so that $\gamma(F_3(S_n))=(\frac{3}{4}+o(1))\binom{n}{2}$, and the upper bound takes the exact form $\gamma(F_3(S_n)) \le \binom{n}{2}-\lfloor (n+1)^2/8 \rfloor$, which is attained for all $3 \le n \le 10$. For $k=4$ we improve the upper bound to $(\frac{17}{27}+o(1))\binom{n}{3}$, which matches the lower bound if Turan's conjecture $\pi_4=\frac{5}{9}$ holds. We show that the packing number of $F_k(S_n)$ equals the maximum size $A(n+1,4,k)$ of a binary constant-weight code, and we characterize when the packing and the domination numbers coincide, namely, for $2 \le k \le n-1$, exactly when $n=2k-1$ and a Steiner system $S(k-2,k-1,2k-1)$ exists; in that case both parameters equal the Catalan number $C_k$.
Read the full abstract
Number Theory
On Rankin Invariants of Ideal Lattices
The Rankin constant $\gamma_{n,k}$ measures the largest discriminant of a densest rank $k$ sublattice of a lattice in $\R^n$ over all such lattices of rank $n$ and their exact values are mostly open. For a general lattice $L\subset \R^n$ of rank $n$, the problem of determining its Rankin invariant $\gamma_{n,k}(L)$ is to find the discriminant of such a densest sublattice. There are no general formulas to compute the discriminant of any arbitrary rank $k$ sublattice. This paper specializes to the case of ideal lattices arising from ideals of number fields, and provides two approaches to study Rankin constants: first, we propose a systematic approach to approximate the minimum discriminant by introducing subideal lattices and proving a closed-form formula to compute $\gamma^{ideal}_{n,k}(L)$, the ideal Rankin invariant of the ideal lattice. Subsequently, we establish necessary conditions on the underlying number field for the existence of an ideal lattice admitting an optimal subideal lattice with a prescribed minimum discriminant. Secondly, we consider ideal lattices with arbitrary sublattices. We improve the current known bounds of the Rankin constants $\gamma_{5,2}$, $\gamma_{7,2}$ and $\gamma_{9,2},\gamma_{9,3},\gamma_{9,4}$, and provide an algebraic construction of the lattice that was used in these dimensions.
Read the full abstract
Machine Learning
TEAM: Topology-Aware, Semantic-Sensitive, General Transformer for Multimodal-Attributed Graphs
Multimodal-Attributed Graphs (MAGs) organize heterogeneous data in graphs whose nodes are characterized by modalities such as text and images rather than traditional unimodal attributes. However, many representative MAG methods organize propagation around node-level states and topology-defined neighborhoods, which can obscure fine-grained modality contributions: a bounded neighborhood may contain semantically weak evidence, while compatible context need not be confined to it. To address both limitations, we propose TEAM, a modality-centric framework that constructs and encodes graph context with node-modality tokens rather than treating each node as a single interaction unit. Concretely, TEAM unfolds the local topology into a Modality Encoding Tree and performs tree-anchored semantic evidence acquisition, organizing local structural context while expanding the accessible evidence space outside the constructed local Tree. We further introduce the Modality-Source Context Transformer (MSCT), which encodes the selected node-modality tokens and coordinates interactions among modality-source groups. By retaining textual and visual attributes as separate interaction units until readout, it supports fine-grained cross-node cross-modal exchange without first collapsing them into whole-node features. Across the evaluated baselines, TEAM attains the highest reported mean on all graph-learning metrics, all three G2Text metrics, and both G2Image metrics. On Toys, it surpasses the strongest baseline by 3.37 and 2.93 percentage points in Accuracy and Macro-F1, respectively.
Read the full abstract
Statistical Methodology
Heavy tails in Gaussian-process surrogates: process-level profiling degeneracy versus observation-level identifiability
In this article, we study heavy-tailed alternatives to Gaussian-process surrogates, namely Student-$t$ processes, $\alpha$-stable processes and Gaussian processes with heavy-tailed noise. These are Gaussian scale mixtures, and the position of the mixing variable decides whether the tails can be learned from a single realization. For process-level mixtures with a free amplitude, the profile likelihood of the mixing law is data-free and is maximized by the Gaussian law, the plug-in predictive is a data-free transformation of the Gaussian-process predictive, and no estimator can be consistent for the mixing law over the model along one path. For observation-level mixtures with symmetric $\alpha$-stable noise, the tail index is identifiable. Further, for one-dimensional inputs under fixed-domain asymptotics with $\alpha_0<2$ and compact hyperparameter sets, a cross-fitted plug-in estimator and the Laplace marginal-likelihood estimate computed at the global posterior mode attain the efficiency bound of the model with a known regression function, also for Matérn-$5/2$ sample paths. We also characterize small-sample failures of the Laplace approximation as folds of a posterior-mode branch and give a diagnostic. Finally, in a repeated Bayesian optimization study, observation-level tails reduced regret under gross outliers, as did Student-$t$ noise, and no degradation was detected under Gaussian noise.
Read the full abstract
Combinatorics
The $(t-1)$-chromatic Ramsey number of cycles for three colours
For integers $t\ge 2$ and $\ell\ge 3$, let $R^t_{t-1}(C_\ell)$ be the least integer $n$ such that every colouring of the edges of $K_n$ with $t$ colours contains a cycle of length $\ell$ whose edges use at most $t-1$ colours. Bucić and Khamseh determined the analogous number for paths and asked for its value for cycles. We answer this question for three colours: we show that $R^3_2(C_\ell)=\ell+\lfloor(\ell+1)/6\rfloor$ for every $\ell\ge 11$, and we determine the values for $3\le \ell\le 10$.
Read the full abstract
Classical Analysis and ODEs
Fefferman-Stein Type Distribution Function Inequalities on Product Spaces
We establish on $R^n\times R^m$ a Fefferman–Stein type distribution function inequality in which the nontangential maximal function of the product Poisson extension is controlled by the Lusin area integral function: \[ |\{N>\lambda\}|\le C\left(|\{A>\lambda\}|+\lambda^{-2}\int_{\{A\le\lambda\}}A^2\,dx\right). \] The central idea is to establish a local $L^2$ estimate for the radial maximal function. This yields a new proof of the corresponding one-parameter inequality, avoiding Green's formula and the treatment of sawtooth regions, and extends to the multiparameter setting.
Read the full abstract
Category Theory
On Proximity Subgroups of Groups
By a $\textit{proximity subgroup}$ of a group $G$ we mean a norm $\nu$ on $G$ valued in a complete lattice $L$ equipped with a binary sum (an additive counterpart of a multiplicative structure). The idea is that the smaller the value $\nu(y)$ is, the closer $y\in G$ is to the subgroup of $G$ defined by the equation $\nu(x)=\bot$, where $\bot$ is the bottom element of $L$. We test whether this notion is a suitable substitute for the usual notion of a subgroup (recovered by taking $L=\{\bot,\top\}$ with join as the sum), by showing that the form of proximity subgroups is noetherian. Intuitively, this means that proximity subgroups are as well suited for formulating and proving isomorphism theorems and homological diagram lemmas for groups as the usual subgroups are. We study several questions in relation to this, including what makes the noetherian form of usual subgroups special among all of these noetherian forms, and in fact, all noetherian forms of the category of groups.
Read the full abstract
Complex Variables
Bott–Chern and Aeppli Cohomology of toroidal groups
We compute the Bott–Chern and Aeppli cohomology groups of theta toroidal groups in arbitrary dimensions, obtaining explicit dimension formulas and bases in all bidegrees. For general toroidal groups, we determine the Hausdorff reductions of these cohomology groups and show that they coincide with those of theta toroidal groups. We also compute the Bott–Chern and Aeppli cohomology groups themselves in certain specific bidegrees, without passing to their Hausdorff reductions.
Read the full abstract
Econometrics
Testing Conditional Moment Restrictions in Networks
Parametric models of network formation, such as dyadic logit models, are increasingly popular, but few tools exist to test their specification. This paper develops a specification test for conditionally independent dyad (CID) models based on the probabilities that these models assign to subgraphs, the patterns of links among small groups of agents. Given the agents' characteristics, a parametric CID model determines not only the probability of each link but also the probability that a group of agents forms a triangle, a star, or any other subgraph. Under correct specification, the expected frequency of each subgraph among agents with given characteristics equals its model-implied probability. This forms the basis of the testable conditional moment restrictions. The test statistic is the weighted integrated squared distance between a nonparametric kernel estimate of these probabilities and the probabilities implied by the estimated parametric model. I derive its asymptotic distribution under the null hypothesis.
Read the full abstract
Mathematical Physics
Exactly integrable field theory in $(2+1)$-dimensional space-time: the spectral equations
The primary object in quantum field theories is the one-step relativistic evolution operator. In this paper we consider the evolution operator defined on the auxiliary Kagomé lattice with the $q$-oscillator algebra of observables, the local ingredient of the evolution operator is the $q$-oscillator $R$-matrix of the Zamolodchikov tetrahedron equation. We construct explicitly the complete system of equations (which can be seen as a specific form of nested Bethe Ansatz equations) defining the spectrum of the evolution operator. This system of equations is valid for an arbitrary initial data (occupation numbers), and the thermodynamical limit is well defined for this system.
Read the full abstract
Optimization and Control
A Multistage Extension of Witsenhausen's Counterexample with Linear Signaling
We study a multistage scalar extension of Witsenhausen's counterexample in which controller 1 observes the initial state exactly, whereas every subsequent controller observes only a local noise-corrupted state and does not inherit its predecessors' observations, so that every nonterminal action both regulates the state and shapes the signal available to the next controller. Restricting the signaling controllers to linear policies while leaving the terminal controller unrestricted, we compute the exact value within this class and reduce the optimization to a deterministic dynamic program in a single propagated variance. The optimal intermediate gains are shown to have a shrinkage form, and the value functions of that program are concave. For two stages, the benchmark admits a closed form on a matching locus, where for small effort weights it has two distinct minimizers, and at the standard parameter instance on that locus a two-point nonlinear policy strictly improves upon it. For three stages, numerical continuation at a reference instance reveals three distinct local minima, and two values of the initial variance at which the lowest of the three changes abruptly. Under the terminal-cost normalization we further prove, for every finite horizon, that the optimal linear value becomes asymptotically equal to the signaling-idle benchmark as the initial variance tends to infinity.
Read the full abstract
Quantum Gases
Coherent Conversion of an Interacting Dark-State-Polariton Condensate into a Photonic Bose-Einstein Condensate
An interacting condensate can retain an observable imprint of its preparation after the interactions are switched off. For light, accessing this regime requires separating the formation of a photon condensate from the material medium that redistributes its energy. We investigate a multimode-cavity scheme in which an interacting dark-state-polariton condensate is coherently converted into a confined photonic Bose-Einstein condensate as the atom-cavity coupling is removed. The polariton mass, matter fraction, and particle number control the condensate occupation during preparation. Remarkably, changing the mass also modifies the condensate's response to repulsion through the spatial overlap of its orbitals, even when the entire single-particle spectrum is held fixed. Ideal mode-preserving conversion conserves the full natural-orbital occupation spectrum, while finite-time few-body calculations show that additional occupation changes during transfer are much smaller than those established during interacting preparation. Once decoupled from matter, the photons evolve under the free cavity Hamiltonian while retaining correlations acquired in the interacting gas. Interference between cavity modes produces oscillations in the photon density, while resonant pair correlations persist after interaction removal and distinguish the resulting photon state from a quadratic thermal state with the same one-body density matrix.
Read the full abstract
Sound
Guidance at Pure Noise Erodes Speaker Similarity in Few-Step Token-Conditioned Flow-Matching TTS
Text-to-speech systems built from a language model and a flow-matching mel decoder now run the decoder with two to four Euler steps instead of the official ten or more. At two steps the word error rate (WER) is unchanged, but speaker similarity (SIM) drops by one to six points. This paper shows that most of this loss comes from classifier-free guidance on the first step. At pure noise the network can only return the conditional mean, so guidance shifts every trajectory by nearly the same vector. Later steps do not remove this shift, and its size grows with the length of the first step, so it is negligible with ten or more steps and accounts for most of the loss with two. Two training-free samplers are derived from this analysis. First-step unguided sampling (FSU) leaves the first step unguided and recovers $84$–$91\%$ of the two-step SIM loss on CosyVoice~2/3 with no significant WER increase. FSU-Mid replaces the two Euler steps by one midpoint step with an unguided first evaluation, recovers $77$–$85\%$ on IndexTTS-2 and matches the ten-step teacher on CosyVoice~2 and GLM-TTS. Both use three network function evaluations, one fewer than the official two-step sampler, and FSU-Mid gives the highest mean recovery among eight published guidance rules re-run at two steps on four decoders.
Read the full abstract
Medical Physics
Deep Learning-based Multiparametric Quantitative Ultrasound for Hepatic Steatosis Assessment
Quantitative ultrasound (QUS) has shown promise for liver fat quantification in metabolic dysfunction-associated fatty liver disease (MAFLD). Incorporating multiple parameters has demonstrated the potential to outperform single QUS parameter measures. Here, we propose a deep learning-based multiparametric QUS framework that learns effective integration of multiple features from QUS measures to estimate pixel-wise hepatic fat fraction, enabling the detecting of heterogeneous fat accumulation within the liver. We developed a U-Net based network architecture to predict local fat fraction percentages, which takes 2D QUS maps as input and is trained with magnetic resonance imaging-proton density fat fraction (MRI-PDFF) as the ground truth. We extracted 2D QUS maps (H-scan frequency, at-tenuation, Nakagami, and RF/envelope/log-compressed data) and tested various input configura-tions consisting of different combinations of the QUS parameters. We enrolled 30 patients with healthy livers or suspected/confirmed MASLD, and acquired ultrasound RF data, and validated our approach using the patient data. Among all tested input configurations, the model combining H-scan, attenuation, and Nak-agami parameters achieved the best performance, with a Spearman correlation coefficient (Rs) of 0.92 and an area under the curve (AUC) of 0.98. The predicted fat fraction maps exhibited color changes concordant with the gradual increasing in MRI-PDFF, from normal liver to 45% fat ac-cumulation. These findings demonstrate that deep learning can integrate multiparametric QUS in-formation in a fully data-driven manner, providing a localized fat fraction map rather than a single whole-liver value. We anticipate that our deep learning framework can offer potential advantages for future assessment of heterogeneous disease development in liver for more complex, progres-sive liver diseases such as steatohepatitis.
Read the full abstract
Artificial Intelligence
When Are Nearby Experts Composable? Cross-Task Model Merging in Neural Thickets
Random perturbations around a shared checkpoint can produce experts for distinct tasks, but shared origin does not ensure that their gains survive composition. We formulate expert-pair selection for joint base-relative retention, measured by strict joint success and worst-task gain with unconditional abstention accounting. An exact finite-difference decomposition and an arbitrarily small orthogonal counterexample show why proximity and orthogonality alone are insufficient. We instantiate this formulation with quality-constrained spectral selection, a fixed-budget framework combining cross-task screening, layerwise pair geometry, separate validation, and Base fallback. Across six checkpoints, 21 task pairs, and five independent perturbation pools, the procedure improves strict joint success over Top-1 averaging throughout and leads the four-method model-scale comparison on four of six checkpoints. For the five models with at least 1.5B parameters, matched quality-only, global-cosine, and spectral ranking attain 49.1%, 50.5%, and 53.1% joint success. Under this matched protocol, the spectral selector also yields the highest worst-task gain and Weak Pareto rate and the lowest regression rate. Mean worst-task gain reaches +0.61 percentage points, while 17.1% regression and 13.0% abstention show that composability remains conditional. Project page: https://anonymous.4open.science/r/Thickets-Merge/.
Read the full abstract
Differential Geometry
A second-order estimate linear in the gradient for complex Hessian equations
We use the maximum principle to prove a second-order estimate with linear dependence on the gradient for the nondegenerate complex Hessian equation on a compact Kähler manifold. The proof combines $C^0$ compactness and a refined concavity inequality with the use of nearby smooth admissible functions. A standard interpolation argument then yields the known uniform gradient and complex Hessian bounds.
Read the full abstract
Quantum Physics
Quantum entanglement as a resource for self-organization and stabilization
Can quantum entanglement generate macroscopic order that persists after the entanglement itself disappears? We establish such a mechanism in a driven-dissipative bosonic lattice, where local entanglement enables transport that amplifies spatial fluctuations into extended Turing patterns. For the full nonlinear dynamics with noise and disorder, we derive rigorous finite-time bounds on pattern intensity for all admissible separable evolutions, allowing arbitrary classical correlations. Quantum dynamics produces pattern intensities beyond these bounds during both formation and stabilization under noise. The excess intensity also provides a lower bound on the accumulated transport contribution beyond the separable limit. We further introduce an environment that responds to emerging order by suppressing local coherence and activating classical transport. The patterns then persist as local entanglement is lost, with both stages arising from a single time-independent Lindblad equation. These results show how transient quantum resources can establish persistent macroscopic order, opening new directions for harnessing self-organization in nonequilibrium systems.
Read the full abstract
Digital Libraries
What Changed in Engineering Writing After Generative AI Chat Assistants: Evidence from 1.8 Million Articles
This study asks what changed in engineering articles after conversational language models became publicly available on November 30, 2022, and develops methods for measuring such change in a field without a uniform abstract index. It assembled 1,839,962 articles published from 2015 to 2025 in 256 engineering journals, tracked twelve measures of wording, style, and article structure, and compared each with the projection of its own pre-2023 trend. Building the monthly series showed that records dated only by year fall on January 1 and create false January peaks unless removed. In abstracts, 31 words associated with language model output appeared 82% more often than projected (55.4 against 27.0 per 10,000 words in 2025), while four control words of the same formal register fell 16%. Words that soften a claim, such as may, fell 13%, words that strengthen one, such as clearly, rose 20%, and long words rose 11%. Author counts, reference lists, page counts, and publication volume stayed within 4% of trend. The change appears in all fourteen subject categories, and false release dates, dropping any one journal or word, and word lists chosen from the data support it. A random sample of 12,866 full texts from the same corpus, including subscription articles and weighted back to it, shows the same words rising 99% in article bodies. Statements disclosing AI writing assistance, measured apart from mentions of language models, went from none to 3.7% in 2025, four in five of them at one publisher. The design is descriptive, not causal.
Read the full abstract
Analysis of PDEs
The Dirichlet problem for complex Monge–Ampère equations with Hölder continuous right-hand sides
We study the Dirichlet problem for the complex Monge–Ampère equation on bounded strongly pseudoconvex domains with $C^3$ boundary and $C^3$ boundary values. For every $n\ge2$ and $0<\alpha<1$, each continuous Bedford–Taylor solution with a strictly positive $C^\alpha$ right-hand side satisfies a global $C^{2,\alpha}$ estimate. There is a unique such solution, which is strictly plurisubharmonic up to the boundary. Combining new boundary estimates with weighted integral estimates for the trace of the complex Hessian along an auxiliary parabolic flow, we obtain global $C^{1,\beta}$ bounds for every fixed $0<\beta<1$ in every dimension. Green function estimates and comparison arguments near the boundary give the boundary Hessian bound. Solutions of auxiliary Dirichlet problems with locally constant right-hand sides have uniform $W^{2,2n^2}$ bounds. Rescaling and interpolation then reduce the global Hessian estimate to interior regularity.
Read the full abstract
Cryptography and Security
Do Benign-Trained Hidden-State Probes Detect Tool Errors Under Adversarial Conditions?
Repeatedly training language models on their own generated data may create a synthetic-data feedback loop in which errors and distributional biases are reintroduced into subsequent training datasets. This paper studies that process in biomedical question answering (QA) using PubMedQA and two Qwen2.5 model sizes, 0.5B and 3B parameters. The study compares a recursive synthetic-data condition, in which generation G(k+1) is trained on answers produced by G(k), against a Human-Control condition that repeatedly uses the original human training data. The study evaluates across four generations from G0-G3 with two random seeds (42 and 123) and a fixed evaluation set of 1,000 expert-labeled samples. The evaluation includes disease and chemical entity F1, context-supported rate, lexical and semantic similarity, answer length, repetition rate, and other evaluation metrics. The Recursive condition for both model sizes and both seeds showed larger declines than the Human-Control condition in disease entity F1, chemical entity F1, context-supported rate, ROUGE-L, and cosine similarity. Under the fixed no-repeat 3-gram decoding constraint, the main observed behavioral change was increased answer length, while the measured 3-gram repetition rate did not increase. The magnitude of the difference-in-change was larger for the 3B model than for the 0.5B model. This difference was particularly apparent in disease F1, context-supported rate, cosine similarity, and answer length. These results show domain-specific changes associated with using recursive synthetic-data training in biomedical QA, but do not establish clinical hallucination rates or universal model collapse.
Read the full abstract
Dynamical Systems
Local study of generic dynamical systems on the b-symplectic plane
Let $\Gb$ be the group of germs of planar diffeomorphisms, fixing the origin, which preserve the standard b-symplectic form. The group acts on itself, under the usual left-right action and we identify, in the generic case, a simple representative of each orbit, depending on a non-vanishing, univariate, function. We then study the action of $\Gb$ on b-symplectic vector fields. For those fields that do not vanish at the origin, we establish a flow-box type theorem, while, for those that do vanish there, we show that the study of their conjugacy classes reduces to a classification of functions. We present this classification, when the origin is a regular point for these functions, while if it is a critical point, we present the orbits of their 2-jets. Finally, we address the problem of whether an element of $\Gb$ can be realised as the time-1 map of a b-symplectic vector field.
Read the full abstract
Machine Learning
ClusLoad: Cluster-Guided Model Sharing for Scalable Distribution Transformer Load Forecasting
Large-scale device-level load forecasting for distribution transformers is challenged by substantial cross-device heterogeneity and the deployment burden of maintaining device-specific forecasting models. Existing approaches typically rely either on device-specific models, which are difficult to scale to large transformer populations, or on global models, which reduce deployment complexity but may overlook heterogeneous load patterns across devices. To address this trade-off, this study develops a trend-clustered forecasting framework that enables model sharing among transformers with similar temporal behaviors. Representation Learning for Time Series using Time-Embeddings (T-Rep) is employed to learn compact temporal representations of transformer load series, which are subsequently grouped using a Gaussian mixture model (GMM). Representative transformers are sampled within each cluster to train cluster-specific Patch Time Series Transformer (PatchTST) models while retaining device-level predictions. The framework is evaluated using real-world 15-min operational data from nearly one thousand distribution transformers. The proposed framework reduces mean absolute error (MAE) by approximately 20% relative to the direct PatchTST baseline, achieves a high-load warning accuracy of 95.66%, while requiring only seven cluster-specific forecasting models. These results demonstrate that behavior-based device organization provides an effective balance between forecasting accuracy, deployment scalability, and risk-oriented operational support for distribution transformer management.