Functional Analysis
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 6 November
A positive $M_n$-valued measure on a locally compact group need not have positive convolution products: matrix multiplication preserves neither self-adjointness nor positivity. We prove that this obstruction disappears completely at an idempotent. If $\mu\in M(G,M_n)$ is positive and $\mu*\mu=\mu$, then all its matrix values commute and there is a unique finite decomposition \[ \mu=\sum_{j=1}^r m_{K_j}\otimes E_j, \] where the $K_j$ are distinct compact subgroups and the $E_j$ are pairwise orthogonal nonzero projections. The result holds for arbitrary locally compact groups, without countability assumptions or a priori commutativity. The proof combines projection-valued dilations, rigidity of contractive idempotent Fourier coefficients, and faithfulness of the regular representation on finite measures over $G\times G$. For normalized positive matrix convolution semigroups, we further identify the Cesàro harmonic projection with convolution by such an idempotent, even if mass escapes to infinity. Its surviving matrix directions are exactly those uniformly confined to a compact subset of the group, yielding an explicit direct Haar-channel decomposition for all $1<p<\infty$ (orthogonal in $L^2$). A bounded Zeno product formula and independent density, atomic, and subgroup-support classifications illustrate the rigidity.
Image and Video Processing
Tomographic Sparse View Selection using the View Covariance Loss
Standard computed tomography (CT) reconstruction algorithms such as filtered back projection (FBP) and Feldkamp-Davis-Kress (FDK) require many views for producing high-quality reconstructions, which can slow image acquisition and increase cost in non-destructive evaluation (NDE) applications. Over the past 20 years, a variety of methods have been developed for computing high-quality CT reconstructions from sparse views. However, the problem of how to select the best views for CT reconstruction remains open. In this paper, we present a novel view covariance loss (VCL) function that measures the joint information of a set of views by approximating the normalized mean squared error (NMSE) of the reconstruction. We present fast algorithms for computing the VCL along with an algorithm for selecting a subset of views that approximately minimizes its value. Our experiments on simulated and measured data indicate that for a fixed number of views our proposed view covariance loss selection (VCLS) algorithm results in reconstructions with lower NRMSE, fewer artifacts, and greater accuracy than current alternative approaches.
Read the full abstract
Number Theory
Harper's beyond square-root conjecture for random matrices
For a Haar unitary matrix $U_N$ of size $N$, we prove that the first absolute moment of $\operatorname{tr}(\operatorname{Sym}^k U_N)$ tends to zero as $\min(N,k)\to\infty$. This settles the random matrix version of Harper's beyond square-root conjecture formulated by Wang and Xu, with no restriction on $k/N$. The proof is elementary and self-contained, combining exact weighted trace moments with a Gaussian change of measure. It also connects the result to the critical chaos model studied by Soundararajan and Zaman, also known as holomorphic multiplicative chaos.
Read the full abstract
Quantum Physics
Universal photonic synthetic temporal-mode converter for heterogeneous quantum interconnects
We propose a synthetic temporal-mode converter (STC) that combines programmable waveform matching with denser time multiplexing for heterogeneous quantum interconnects. Programmable couplers and optical delays reshape known single-photon temporal waveforms through coherent interference between time bins. Conversion to a common single-bin mode concentrates amplitudes from $N$ bins in $\lceil\log_2N\rceil$ mixing stages, simultaneously improving temporal indistinguishability and freeing channel time. Simulations yield mode overlaps above 99.9% for exponential and Gaussian inputs. For $N=256$ and $0.5\,\mathrm{dB}$ loss per stage, STC gives a 45.6-fold higher photon-pair coincidence rate than temporal gating at equal mode overlap for two exponential inputs. Relative to long pulses, the shortened temporal matched mode can support an approximately 41-fold higher photon-pair coincidence rate when the available time-multiplexing capacity is fully utilized.
Read the full abstract
Mathematical Finance
Drift-invariant kernels for discount curve models
Kernel ridge regression estimates the discount curve as the minimizer of a pricing error plus a penalty given by the norm of a reproducing kernel Hilbert space. By the representer theorem, the estimate lies in a finite-dimensional affine space that depends only on the kernel and on the cash flow dates of the instruments. We ask when this space is invariant under the risk-neutral drift of the discount curve, so that the estimated curves can be the states of a continuous-time arbitrage-free model. We show that this holds for every grid of cash flow dates if and only if the kernel is of the form $k(x,y)=\e^{b(x+y)}\big(c\,\e^{axy}+1-c\big)$, apart from kernels whose Hilbert space has dimension at most two. For this kernel the estimated curve is an affine combination of flat discount curves, each corresponding to a constant continuously compounded yield. The drift-invariant subspaces correspond to sub-grids of the cash flow dates, and the discounted coefficients have linear risk-neutral dynamics.
Read the full abstract
Optimization and Control
Attractivity Implies Lyapunov Stability for Multidimensional Fornasini–Marchesini Systems
We study the linear multidimensional Fornasini–Marchesini recurrence with uniformly bounded boundary data prescribed on the open coordinate faces. Attractivity is understood in the sense of Pringsheim convergence, in which all discrete indices tend to infinity. We establish that attractivity alone implies Lyapunov stability, and consequently is equivalent to asymptotic stability, for an arbitrary number of discrete directions. Our proof combines inheritance of attractivity by lower-dimensional subsystems, a Baire-category uniform boundedness argument, and word-sum estimates to control coordinate slabs outside the Pringsheim tail. We use a generalized spectral stability criterion proved in a separate companion preprint, without reproving that criterion here. We also obtain a uniform exponential interior estimate and contrast our boundary/convergence convention with an earlier two-dimensional attractivity theorem.
Read the full abstract
Rings and Algebras
Division ring extensions and a counterexample to the pure semisimplicity conjecture
We construct a family of division ring extensions and compute their left and right dimensions. In characteristic zero, we obtain a counterexample to the pure semisimplicity conjecture: a hereditary right pure semisimple ring of infinite representation type. For each prime $p\ge3$, we construct in characteristic $p$ a hereditary artinian ring of finite representation type with Coxeter diagram $I_2(p+2)$.
Read the full abstract
Populations and Evolution
Predicting fitness from genomic offset: a multivariate latent-trait framework
Genomic offset methods quantify the genetic consequences of environmental change using genotype environment associations (GEA), but the conditions under which they predict fitness remain poorly understood. Here, we develop a theoretical framework linking genomic offset to relative fitness loss through multivariate latent adaptive traits under stabilizing selection. We show that squared genomic offset is linearly related to expected log relative fitness loss when averaged over genetic architectures, provided that direct environmental effects are absent. However, this relationship does not necessarily hold for a realized genetic architecture. We identify two complementary mechanisms that can explain high predictive accuracy of genomic offset statistics. First, when effective adaptive dimensionality is sufficiently large, the genetically mediated fitness response concentrates around its marginal expectation, reducing the dependence on the particular genetic architecture. Second, even with few adaptive traits, accurate predictions can arise from geometric alignment between the complete fitness loss matrix and the GEA metric. This second mechanism also encompasses direct environmental effects and their interactions with genetically mediated responses, which may preserve or weaken predictive accuracy depending on their alignment. We further derive bounds on predictive accuracy that characterize these effects. Our results clarify the assumptions and limitations underlying genomic offset predictions and suggest that genomic offset is best interpreted as a measure of genetically informed adaptive mismatch, rather than a direct predictor of realized fitness or any particular phenotypic trait.
Read the full abstract
Data Structures and Algorithms
Contention Resolution Against an Adversarial Scheduler
In the classical contention-resolution problem, a set of players share a broadcast channel, and a player succeeds in a given time step if it is the only player to broadcast during that step. Implicit in the classical problem is the assumption that every player is free to act on every step: once a player arrives, it participates until it succeeds. We study what happens when that assumption is removed. In our model, an adversarial scheduler decides, on each step, which subset of the players gets scheduled to participate in the step. An unscheduled player simply sleeps through the step. We measure a protocol by its efficiency: the number of players divided by the expected number of steps in which at least one player was active. Our first result says that, so long as the players are given access to a global clock and to shared randomness, they can achieve efficiency $\Omega(1/\log n)$. The players are able to achieve this bound without requiring any channel feedback, and even if the adversarial scheduler is adaptive. We also give a matching lower bound of $O(1 / \log n)$, which holds even with channel feedback, and even for an oblivious scheduler. Our upper bounds all require the players to have access to a shared global clock and to shared randomness. Our final result says that this type of implicit shared state is necessary. If the players have neither a global clock nor shared randomness, then no protocol can achieve efficiency better than $O(\log n / n)$; and this holds even if the other aspects of the problem are designed to be as favorable for the players as possible (the channel gives ternary feedback and the adversary is oblivious).
Read the full abstract
Artificial Intelligence
Continuous wavelet-based feature extraction for laser-induced breakdown spectroscopy classification of human plasma samples
Extracting discriminative features from laser-induced breakdown spectroscopy (LIBS) spectra is important for the classification of complex biological samples. This study evaluates continuous wavelet transform (CWT)-based spectral peak extraction for distinguishing human plasma samples labeled as SARS-CoV-2-positive or negative. Plasma specimens from 181 individuals, comprising 53 positive samples and 128 negative controls, were deposited as 5 {\mu}L droplets on silicon substrates and dried before analysis. Spectra were acquired using a nanosecond Nd:YAG laser operating at 1064 nm, with 100 single-shot measurements collected per specimen and averaged after intensity-based outlier rejection. CWT-based peak detection was coupled with a support vector machine (SVM) classifier. The reported leave-one-out evaluation yielded an accuracy of 93.00%, sensitivity of 94.67%, and specificity of 91.61%; these estimates remain provisional pending reconciliation with specimen-level predictions. Features identified by the CWT pipeline included emission lines assigned to Ca II, K I, and N I. The workflow provides a basis for examining spectral features associated with the studied plasma groups. Independent validation and assessment of specimen, substrate, and acquisition-related variability are required to establish generalizability and diagnostic relevance.
Read the full abstract
Metric Geometry
Multiscale Closing Index (MCI) for measuring structural similarity between 2D point sets
We introduce Multi-scale Closing Index (MCI), a new dissimilarity measure for evaluating the alignment of two 2D point sets. MCI is bounded between 0 and 1 and is based on multiscale mathematical morphology operations, with smaller values indicating greater structural agreement. The measure was motivated by the need to assess the quality of registration between point sets representing cell populations in biological images. To evaluate its behaviour under controlled conditions, we generated synthetic point clouds from Perlin-noise-based spatial structures, following established approaches for simulating biologically plausible spatial organization. We assessed MCI between those point clouds and different transformed versions, obtained by local perturbation, point removal, translation, rotation and addition of impulsive noise. Across these experiments, the dissimilarity MCI increased gradually and consistently with perturbation severity, while showing a clear discontinuity between clouds originating from the same underlying spatial structure and point clouds generated from different structures. We further compared MCI with commonly used point-set distances, including Hausdorff-based measures and Sliced Wasserstein distance. These experiments show the ability of MCI to capture structural agreement and quantify deviations between point sets, highlighting its potential as a complementary metric for evaluating two-dimensional point-set registration.
Read the full abstract
Software Engineering
QAOA for Layered Architecture Recovery: Feasibility, Limits, and the Role of Decomposition
Layered architecture recovery (LAR) reconstructs a system's architectural layers from its package dependencies, and is defined as a combinatorial optimisation problem solved with classical search. We investigate whether the Quantum Approximate Optimization Algorithm (QAOA) can be applied to LAR, and with what practical limitations. We formulate LAR as a Quadratic Semi-Assignment Problem, transform it into a Quadratic Unconstrained Binary Optimisation (QUBO) problem in which each package is penalised unless it is assigned to exactly one layer, and encode it as an Ising Hamiltonian solved by QAOA. To reach instances beyond the qubit budget of a single circuit, we also apply an impact-guided decomposition that solves the problem as a sequence of smaller sub-problems. We evaluate the approach on three real Java systems ($15$, $18$, and $24$ qubits), studying the effect of circuit depth and sampling budget (RQ1), a comparison with a classical Tabu Search baseline (RQ2), robustness to simulated quantum noise (RQ3), and the decomposition (RQ4). We find that monolithic QAOA is feasible and accurate on the two smaller systems but fails on the $24$-qubit instance, that the sampling budget rather than circuit depth governs its accuracy, and that it is competitive with but does not outperform Tabu Search. Decomposition restores reliable recovery on the $24$-qubit instance, raising the optimal rate from $6.7\%$ to $93.3\%$; its accuracy is governed by the size of each sub-problem and, unlike the monolithic pipeline, improves as the sampling budget is reduced. Our results indicate that, in the current noisy intermediate-scale regime, problem decomposition, rather than larger circuits or more sampling, is the key to applying QAOA to software architecture recovery.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Does DESI point towards transient cosmic acceleration?
We investigate an oscillatory equation of state for dark energy that oscillates around the mean value $-1$ and includes a source term proportional to the cube of the number of e-folds. Due to this source term, the oscillation amplitude shifts towards the $w_{\rm de}<-1$ regime at higher redshifts, whereas in the future, the source term drives the oscillation amplitude towards the $w_{\rm de}>-1$ regime. As the equation of state enters the $w_{\rm de}>-1$ regime in the future, the dark energy density dilutes, causing the Hubble parameter to decrease asymptotically towards zero, so that the present acceleration is transient: the Universe transitions to decelerated expansion and approaches a static state as $a\to\infty$, without ever contracting. We compare this equation of state with BellDE, CPL and flat $\Lambda$CDM. We confront the model with a combination of datasets, including Planck, DESI BAO and three different supernova compilations: Pantheon+, Union3 and DES-Dovekie. We compare the log-evidence of each model with that of $\Lambda$CDM. The dynamical dark energy models yield $H_0$ values in the range $66.0^{+0.80}_{-0.79}$–$67.84^{+0.69}_{-0.74}\,{\rm km\,s^{-1}\,Mpc^{-1}}$, with matter density values in the range $\Omega_{\rm m}=0.3067^{+0.0073}_{-0.0072}$–$0.3271^{+0.0083}_{-0.0080}$ across all the combinations. These values are consistent with those of the baseline model, apart from a mild upward shift in $\Omega_{\rm m}$ for the Union3 combination. Among all the models, BellDE yields comparatively stronger Bayesian evidence than the other models, with weak evidence for the Pantheon+ compilation and moderate evidence for the Union3 and DES-Dovekie compilations. The sourced oscillatory model yields inconclusive evidence for Pantheon + and Union3, while weak evidence for DES-Dovekie relative to $\Lambda$CDM.
Read the full abstract
Physics and Society
Regime change in a model for political opinion dynamics
Political crises involving a series of violent events are the rule rather than the exception in history. Here, we use a minimal opinion dynamics model to describe regime change as a contested transition between dominant shared ideas in a society. New opinions spread with a cumulative advantage, and agents do not return to opinions they held before, resulting in recurrent boom-and-bust dynamics without tuned parameters. During a transition, opinion changes grow approximately exponentially and then saturate in the finite system, declining approximately linearly as spreading is limited by the shrinking interfaces between opinion groups. We compare qualitative features of these transitions with historical data of political violence, used as an empirical proxy for contested transitions. A permutation test shows that violent events cluster in time in the dense Chinese and US records. Around their peaks, violent events rise in an accelerating way and then decline; in the Chinese data, the decline is consistent with the approximately linear decline in the model. We also use a fragmentation measure that serves as an early warning signal for transitions in the model. Overall, the patterns in the empirical data and the dynamical behavior of the model could contribute to a better understanding of long-term historical processes.
Read the full abstract
Social and Information Networks
Research Collaborations Within and Across Disciplines in NIH-Funded Atlas Programs
The National Institutes of Health (NIH) fund several programs (aka consortia) that support the construction of organ-specific or whole person atlases. These initiatives are designed to build comprehensive open reference datasets while advancing knowledge graphs, analytics, visualizations, and user interfaces needed to explore multiscale human atlas data. Here, we compare the Human Tumor Atlas Network (HTAN), Human BioMolecular Atlas Program (HuBMAP) and Cellular Senescence Network (SenNet) to examine collaboration by consortium component (CC) and researcher role and its relation to scientific impact. We retrieved 884 NIH award records and identified 1,255 focal publications acknowledging these awards, involving 8,952 researchers; compiled 99,108 citing publications covering 93% of the focal publications; constructed co-authorship networks; and as single-project publications or internal collaborations (IC), including intra-CC and inter-CCs collaborations. within each atlas program. Results show that IC publications are rare but involve many funded projects and principal investigators (PIs). Technology-oriented CCs tend to engage in intra-CC collaboration, whereas experimental-oriented CCs often collaborate across CCs of the same type. Compared with other researchers, PIs exhibited a significantly stronger association with inter-CCs than with single-project publications. IC publications have higher mean Relative Citation Ratios than single-project publications. Inter-CCs publications have significantly higher cumulative and annual citation levels in the first five post-publication years. Study methods and results can be used to inform the funding, management, and impact reporting for future Big Science programs that bring together 300+ experts from different disciplines to address major research challenges.
Read the full abstract
Computer Vision
Zero-Shot Benchmarking of Unified Multimodal Models for Shallow-Water UAV Image Distortion Assessment and Restoration
Shallow-water UAV imagery provides centimeter-scale observations of benthic environments, but water-induced distortions complicate interpretation. To our knowledge, we present the first systematic zero-shot benchmark of unified multimodal large language models (MLLMs) for both assessing water-induced distortions and restoring shallow-water imagery within the same model framework. We evaluate five open-weight unified MLLMs with frozen pretrained weights and without access to paired reference images or descriptions of the true seabed. The benchmark contains 1,000 image pairs: 600 semi-synthetic UAV scenes created by applying synthetic water distortions to real shallow-water imagery and 400 fully synthetic through-water scenes. Three independent human annotations cover five distortion types. We evaluate four assessment prompts (P0-P3), from direct instruction to structured task decomposition, on all 1,000 images and four diagnosis-guided restoration prompts (R0-R3) on a common 100-image subset. Assessment varies substantially by model and prompt, with systematic prediction biases and a best pooled balanced accuracy of 0.654 for distortion presence. The best pooled PSNR gain is +0.255 dB, no unified MLLM improves pooled LPIPS, and some outputs alter scene structure or hallucinate content. Two restoration-only baselines also show limited transfer: InstructIR achieves a pooled PSNR gain of +0.040 dB, while PromptFix degrades all four pooled metrics under every prompt. These results show that recognizing shallow-water degradation does not imply reliable physical restoration and provide a reproducible baseline for future work.
Read the full abstract
Spectral Theory
Finding geometric criteria for negative eigenvalues of the Neumann–Poincaré operator
Motivated by a question of Grigori Rozenblum, we study how the geometry of the boundary of a convex domain $\Omega\subset\mathbb{R}^3$ affects whether the Neumann–Poincaré operator on $\partial\Omega$ has negative eigenvalues. First, we show that a flat face on the boundary of a bounded, convex Lipschitz domain produces an infinite-dimensional negative spectral subspace. For convex domains with $C^{1,\alpha}$ boundary, we bound the number of negative eigenvalues from below by the Dirichlet eigenvalue counting function of a projected patch over which the chords are sufficiently short. A Rellich identity gives a sufficient condition on a defining function to exclude negative eigenvalues. We also report numerical computations, which led us to the theoretical results.
Read the full abstract
Quantum Physics
Witnessing Polarization Entanglement of Annihilation Photons via Compton Scattering
Experimentally certifying polarization entanglement in annihilation photons is challenging because polarizers are unavailable in the high-energy regime and one needs to rely on Compton polarimetry. Previously proposed schemes designed for this task are either experimentally unfeasible or require a specific source state model. We address these issues by reformulating Compton scattering as a continuous-outcome quantum instrument, thereby enabling the prediction of the scattering angles and the post-scattering polarization state. Our framework allows connecting Compton polarimetry with entanglement witnesses by deriving experimentally accessible lower bounds on Bell state fidelities for arbitrary initial two-photon polarization states. We discuss how this witness can be extracted from coarse-grained coincidence measurements with detectors of almost any angular resolution. The presented scheme does not require knowledge of angle-independent empirical factors like detector efficiencies and source activity. Thus, our results provide a feasible route towards experimental entanglement certification in para-positronium decay with current technology.
Read the full abstract
General Relativity and Quantum Cosmology
Scalar Black-Holes in Gravity from Entropy
We investigate static, spherically symmetric black holes in Gravity from Entropy (GfE), focusing on the sector in which the bosonic form-space field is a pure scalar. Starting from the full auxiliary-field equations, we show that the Dirac–Kähler matter equation contains both a scalar equation and a two-form consistency condition. The latter is automatically satisfied by a radial scalar. We derive an exact integral identity, valid for a general backreacted static spherical geometry, and prove that a regular scalar approaching zero at infinity must vanish if the sampled scalar and radial auxiliary-field eigenvalues are positive and the effective mass squared $m^2+\xi R$ is nonnegative. In particular, this excludes minimally coupled static scalar hair with $m^2\geq0$ on the regular identity-connected branch. We also present the perturbative GfE vacuum geometry and use its curvature to identify the region in which nonminimal coupling can evade the sign argument. A second, sharper estimate shows that making $m^2+\xi R$ negative is not by itself sufficient to produce a regular static zero mode: a finite interval of nonminimal couplings still admits no branch bifurcating from the vacuum solution.
Read the full abstract
Machine Learning
Safety Is $\Pi_1$, Provability Is $\Sigma_1$: One Framework for the Hugging Face Breach and OpenAI's Mathematical Proofs
OpenAI's agents breached Hugging Face in July 2026, and the same company released more than seven hundred AI-generated manuscripts of mathematical proofs in October 2026. The first event is read as a failure in security, while the second is read as a success in mathematics. We read both through the hybrid framework of our paper (Kim and Kang, 2026, arXiv:2611.01914), in which coding, proving, and learning from data are derivations by one and the same deductive mechanism, distinguished by the kinds of premises from which they are derived. Under this framework, the AI systems behind the two events ran the same loop, in which a statistical proposer produces candidate exploits or proofs and a deductive verifier settles each one by a derivation. What separates the events is the logical form of the statement wanted by each of the two domains, security and mathematics. The safety of a program is a $\Pi_1$ statement, for it says that no exploit exists, which is not assured by any finite list of tests that have been passed by the program. By contrast, the provability of a sentence is a $\Sigma_1$ statement, for it says that a proof exists, which is assured by a single instance, namely a proof that has been checked by the deductive verifier. The theorem behind both is Turing's Undecidability Theorem: the halting set is $\Sigma_1$ but not decidable, so its complement is $\Pi_1$ and not $\Sigma_1$, which is to say that membership has a finite witness but non-membership has none. We prove that attaching a statistical proposer to a deductive verifier does not change this undecidability result, however good the proposer is. We then discuss what does assure safety and where the residual risk goes.
Read the full abstract
Networking
5G-NetBotSim: A Synchronized Co-Simulation Framework for Connected Robots over ROS and srsRAN-based 5G Network
Connected robotic applications increasingly rely on wireless communication for teleoperation, coordination, and distributed autonomy. However, robotic simulators often use ideal links or abstract delay models that do not capture cellular scheduling and protocol processing effects. We present 5G NetBotSim, a realtime co simulation framework integrating ROS/Gazebo workloads with an executable, full stack srsRAN 5G network emulator. The framework provides isolated User Equipment (UE) robot contexts and stage level timing across the host, gNB, UE, ROS, and robot execution stages. We evaluate human to robot teleoperation and robot to robot leader follower coordination under controlled channel quality conditions. Results show how CQI and MCS variations propagate to throughput, communication latency, and robot performance. In leader follower coordination, higher channel quality reduces mean tracking error from 0.558 m to 0.460 m, a 17.6% reduction. These results demonstrate protocol stack in the loop co simulation for joint network and robotic-application evaluation.
Read the full abstract
Cryptography and Security
GuardrailAgent: An Agentic Framework for Jailbreak Defense
Despite undergoing safety training to align with human values, large language models (LLMs) remain vulnerable to jailbreak attacks that circumvent built-in safeguards and elicit malicious behaviors given by the adversary. Existing defenses are mostly static and fail to adapt to new or unknown attack strategies, providing limited resilience. To address these, we propose a multi-agent defense framework, GuardrailAgent, that offers adaptive, end-to-end protection. The system autonomously gathers and summarizes attack characteristics into a continuously updated knowledge base that supports both input- and output-side detection. A feedback module integrates misclassified cases for iterative improvement at minimal cost. Tested on over ten representative jailbreaks, the framework achieved a 93.1% input-side detection rate, 93.9% overall detection, and a 95.8% F1 score, while rapidly adapting to unseen attacks. These results demonstrate GuardrailAgent's outstanding capability to maintain LLM safety in the face of evolving jailbreak tactics.
Read the full abstract
Computational Complexity
Metric Dimension Parameterized by Bag-Connected Treewidth and Bag-Connected Pathwidth
The metric dimension of a graph is the minimum size of a set of vertices whose distance vectors distinguish all vertices of the graph. We prove that \MetricDimension{} is NP-complete on planar graphs of treewidth and bag-connected treewidth exactly three. The graphs in our construction have no induced cycle longer than four and have (prefix-connected) pathwidth at most four. In particular, the problem admits no XP algorithm parameterized by bag-connected treewidth unless $\mathrm P=\mathrm{NP}$. This answers questions of Bonnet and Purohit and of Bousquet, Deschamps, and Parreau. We prove that the problem is fixed-parameter tractable when parameterized by bag-connected pathwidth.
Read the full abstract
Probability
Strong Convergence for Zig-Zag Products and Typical Spectra of Irregular Graphs
We investigate the following: do existence bounds obtained by interlacing methods describe the spectra of typical random graphs? Depending on the model, we provide both positive and negative answers. We establish a Friedman-type theorem for zig-zag products of random regular graphs $G$ with a fixed vertex-transitive gadget $H$, where $G$ is sampled from independent permutations or matchings. We prove that the new spectrum converges in Hausdorff distance to that of the zig-zag product of the infinite regular tree with $H$. In particular, the new spectral radius converges to the sharp threshold established by Cohen, Cohen, and Maor. We establish the spectral limit using strong convergence and, drawing on quantitative estimates of Chen, Garza-Vargas, Tropp, and van Handel, obtain an $O(1/n)$ upper-tail bound in the permutation model. For the irregular configuration model, we show that the interlacing bound of Cohen and Maor, asymptotically $2\sqrt{\mu-1}/\mu$ for limiting average degree $\mu$, is atypical: for uniformly bounded degree sequences with a nonconstant limiting degree law of minimum $\delta\ge 2$, a random graph exceeds the interlacing bound by a positive constant with high probability. This gap is explained by the limiting spectral distribution, whose support edge is $2\sqrt{\delta-1}/\delta$. Strikingly, unlike the familiar behavior of random regular graphs, our experiments reveal second eigenvalues that remain far below this limiting edge even for the largest graphs tested. For $\delta\ge3$, we propose that the scarcity of minimum-degree trees explains this behavior; estimates for their size suggest spectral witnesses within $O\left((\log\log n)^{-2}\right)$ of the limiting edge, which is consistent with our numerical observations.
Read the full abstract
Functional Analysis
Toeplitz operators with bounded harmonic symbols on weighted Bergman spaces are the fixed points of a normal unital completely positive map
In this article, we consider weighted Bergman spaces over open unit disk corresponding to nonnegative integrable radial functions with positive mass arbitrarily close to the unit circle. Then we show that Toeplitz operators with bounded harmonic symbols on the weighted Bergman space are the fixed points of a normal unital completely positive map. This result generalizes the classical characterization of Toeplitz operators on the Hardy space due to Brown and Halmos. Moreover, we prove that there is no compact Toeplitz operators with bounded harmonic symbols on weighted Bergman space.
Read the full abstract
Artificial Intelligence
What to Delegate, What to Preserve: Human Preferences for Agentic AI Across Everyday Life and Work
AI agents can execute multi-step tasks on users' behalf, yet the boundaries of acceptable delegation remain unclear. We report a scenario-based survey of 289 U.S. adults with prior AI experience covering 45 use cases across nine life domains. Each respondent evaluated one scenario per domain, producing repeated measures of automation preference, trust, collaboration, AI leadership, privacy concern, and willingness to pay. Automation preferences varied substantially across domains. Generalized estimating equation models showed that greater trust in the AI agent, desired AI leadership, prior task-specific AI use, and anticipated financial savings were associated with greater automation preference. In contrast, concerns that automation would reduce personal fulfillment predicted lower automation preference. Open-ended responses further revealed a preference for bounded autonomy, with participants often favoring substantial AI execution while retaining human involvement, oversight, or final authority. We argue that delegation is best understood as a configuration of execution, collaboration, leadership, and authority.
Read the full abstract
Machine Learning
One Run Is Not Enough: A Monte Carlo Protocol for Evaluating LLM Reliability on Data Science Tasks
Large language models (LLMs) have shown promise for automating data science workflows. Benchmarks such as DataSciBench (2026) and StatQA (2024) have advanced the evaluation of their capabilities across data science and statistical tasks. Assessing their reliability requires examining both the quality of their responses and their consistency across repeated runs, task types, and difficulty levels. With this aim, we introduce DS-QA, a benchmark comprising 100 data science and statistics questions drawn from UK university courses, spanning 16 datasets and three difficulty levels. We develop a three-layer evaluation framework that combines general response properties (runtime, verbosity and text similarity), coarse-grained metrics (accuracy, code executability and completeness), and task-specific metrics. We evaluate OpenAI models using a Monte Carlo protocol with 100 repeated runs. The results reveal substantial variability in responses to data visualisation and supervised learning tasks, particularly at medium and hard difficulty levels, highlighting the importance of evaluating consistency alongside response quality. We also find that an increase of 0.1 in the word verbosity ratio corresponds to an increase of about 0.013 in the token verbosity ratio.
Read the full abstract
Information Retrieval
SAGE-PAIR: Source-Pretrained Routed Residual Adaptation for Sparse-Feedback Recommendation
Recommendation systems often enter a new serving window before enough outcomes are available to retrain the deployed model. We study this regime with 0.05-1.0% target feedback, a range that spans our stress tests while keeping supervision far below the degrees of freedom of the source recommender. At this scale, existing fine-tuning and meta-learning methods still estimate target-specific updates from extremely sparse feedback, leaving the update direction weakly identified. SAGE-PAIR instead transfers a source-supported residual correction space and only calibrates this space using target outcomes. We introduce SAGE-PAIR, a staged adapter that pretrains a low-dimensional global residual on an adjacent source period and adds a soft router over residual experts conditioned on source user profiles. Target feedback then calibrates this residual state rather than relearning the recommender. On a locked test split of KuaiSAR-v2, SAGE-PAIR (Full) achieves a 1.77% relative improvement in NDCG@10, while a core-protocol comparison shows a further 0.25% gain from routed experts beyond source-pretrained PAIR. Matched source-pretrained LoRA and Adapter-Tuning baselines remain near the source model, supporting the need for a different adaptation space. Thus, SAGE-PAIR improves ranking quality with millisecond-scale adaptation and inference costs. Sparse-feedback adaptation is fundamentally governed by the identifiability of the update space rather than by label count alone. Source pretraining supplies estimable correction directions for scarce labels, while user-conditioned routing contributes a consistent, differentiated cross-user correction on top of the global correction. Our code is available at https://anonymous.4open.science/r/SAGE-PAIR-610E/.
Read the full abstract
Physics and Society
Scarce or Missed Opportunities? Structural Peers and Latin America's Diversification Frontier
Latin America's limited export upgrading is often attributed to failures to seize available diversification opportunities. We distinguish two possibilities: countries failed to realize opportunities demonstrated by structurally similar economies, or upgrading-oriented options were scarce within this peer-revealed set. Using export data for 105 countries from 1962 to 2024, we construct country-specific opportunity sets with an imitation-density measure associated with subsequent product entry. Latin America's deficit arose mainly from scarce upgrading-oriented candidates rather than weak aggregate conversion. Conversion rates matched the global benchmark, but regional frontiers contained few manufacturing products and almost no higher-complexity manufacturing candidates and, unlike East and Southeast Asia, did not shift toward them over time. An illustrative decomposition associates most of the regional upgrading-entry gap with candidate availability. The findings imply that industrial policy should expand feasible upgrading pathways rather than focus only on converting existing opportunities.
Read the full abstract
Probability
Extreme-value asymptotics for smallest enclosing balls of spherically symmetric samples
Let $R_n$ denote the radius of the smallest ball containing an i.i.d.\ sample of spherically symmetric random vectors in $\mathbb R^d$, and let $D_n$ be the sample diameter. While the asymptotic behaviour of $D_n$ is well understood, much less is known about $R_n$. Starting from a classical result of Daniels for planar Gaussian samples, we develop a point-process approach to the asymptotic behaviour of $R_n$ in the three classical extreme-value regimes. We obtain limit theorems in the Gumbel, Fr'echet and Weibull cases, with the limiting laws described by geometric functionals of Poisson point processes. In dimension two, the Gumbel limit can be identified explicitly and is universal within the corresponding max-domain of attraction. We also study the asymptotic relationship between $R_n$ and $D_n$ and the supporting configurations of the smallest enclosing ball.