Numerical Analysis
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 13 September
A tensor randomized average block coordinate descent method with heavy-ball momentum is proposed for the tensor least squares problem with respect to the t-product. Theoretical analysis for tensor block coordinate descent method is established and average techniques are applied to avoid the computation of tensor Moore–Penrose inverse. To further accelerate convergence, a heavy-ball momentum scheme is incorporated into the block coordinate descent method, where the step size and momentum parameters are adaptively determined by a two-dimensional minimal residual projection. Theoretical analyses give the convergence of the new method and provide an improved bound on the linear convergence rate. Numerical experiments further verify the efficiency of the proposed method in terms of the number of iterations and the better video recovery performance.
Metric Geometry
Odd branching obstructs multiplicity-one integral current structures
Giuliano Basso asked whether every $n$-dimensional integral current space admits an integral current structure with the same characteristic set and multiplicity one almost everywhere. We give a negative answer already in dimension one. The obstruction is a parity phenomenon at odd-valence branch points. More precisely, if a one-dimensional integral current has unit multiplicity on the arms incident to an isolated vertex $v$, then the coefficient of its boundary at $v$ has the same parity as $\deg(v)$. Hence infinitely many odd-valence vertices force infinite boundary mass for every full-support unit-multiplicity current. We construct a compact geodesic, doubling, $1$-Ahlfors regular graph-like continuum $G$ of finite length and a boundaryless current $T\in\I_1(G)$ with $\set(T)=G$, while no $S\in\I_1(G)$ with $\set(S)=G$ can have multiplicity one $\Hh^1$-almost everywhere. We classify all full-support integral cycles on $G$. In particular, the least possible essential maximal multiplicity is exactly two, and the minimum mass of such a cycle is $3/2$. For finite metric graphs we identify a complementary sharp defect: the minimum boundary mass among unit-multiplicity currents equals the number of odd-degree vertices. Finally, by taking products with flat tori and using the precise Ambrosio–Kirchheim slice representation, we obtain boundaryless compact geodesic Ahlfors-regular counterexamples in every dimension.
Read the full abstract
Artificial Intelligence
OptoAgent: A Trustworthy Multi-Agent Framework for Opportunistic Vision Micro-Screening in Classroom Environments
A child with reduced distance vision often does not know that anything is wrong. Children adapt, move closer, and rarely report the difficulty, so the problem can survive years of schooling before an adult notices. School screening addresses part of this, but it runs on a schedule, depends on staffing, and is separated from the classroom moments where the difficulty appears. Smartphone and web-based acuity tests have widened access, yet every one of them still needs somebody to start a test. We present SightSentinel, an architecture that turns a wall display a child already reads from into a recurring screening site. Ordinary educational content carries short calibrated optotype probes, and eight specialized agents divide the work. Perception agents recover viewing distance, recognition accuracy, approach behavior, gaze stability, response latency, and interocular difference from each encounter. A quality agent discards observations taken under bad geometry, poor lighting, or inattention. A longitudinal agent accumulates only the surviving evidence against the child's own baseline, and an orchestrator reports a Vision Concern Score routed through a safety gate whose output range excludes diagnosis, refraction, prescription, and reassurance. The design question is whether many cheap, noisy, well-gated encounters can reach a referral decision that one scheduled test reaches late or misses. We state the formulation, the architecture, a four-stage validation protocol against clinical reference standards, and the conditions under which the approach should be rejected.
Read the full abstract
Analysis of PDEs
Sharp comparison between the perimeter and its fractional analogue in two dimensions
We show that the fractional perimeter of any order $0<s<1$ is maximal, among all planar sets of fixed perimeter, for the disk.
Artificial Intelligence
Beyond Scene Description: Multi-Agent Orchestration for Non-visual Access to Virtual Worlds
Virtual worlds now host classrooms, meetings, conferences, shops, and social venues, and nearly every interaction they expose assumes a user who can scan a three-dimensional scene, follow avatars, and read floating panels. Blind and visually impaired (BVI) users are left with assistive tools that each solve one task in isolation: naming an object, reading text, describing a scene, or planning a route. A live virtual room defeats that model: obstacles, speakers, gestures, chat, slides, and notifications arrive together, and a tool that narrates all of them trades a visual barrier for an auditory one. This paper presents MetaBlind, an architecture that distributes nonvisual access across eight specialized agents, spanning perception, navigation, social and object interaction, communication, safety and trust, memory, and personalization, and that places an Accessibility Orchestrator between those agents and the user. Agents publish candidate information into a shared accessibility context instead of speaking to the user directly. The orchestrator scores each candidate on safety relevance, goal relevance, urgency, confidence, user relevance, and estimated listening load, then releases only the items it judges relevant at that moment through speech, structured audio, or haptic output. We give the selection step a formal statement, specify the orchestration cycle as an algorithm, and define an evaluation protocol against a single-agent assistant. MetaBlind is reported at the design stage, with no prototype measurement or user study, and the protocol states which outcomes would support the design and which would refute it.
Read the full abstract
Signal Processing
Combating Suppressive Jamming with Dynamic Agile Reconfigurable Intelligent Surface Antenna Array (DARISAA)
Suppressive jamming is a severe challenge to digital receivers in wireless communications, since high-power jamming may cause analog-to-digital converter (ADC) overload or automatic gain control (AGC)-limited quantization blocking of weak desired signals. Analog beamforming in the spatial domain can be employed to combat suppressive jamming signals before they reach the ADC. In this paper, we introduce a comprehensive anti-jamming scheme considering both direction-of-arrival (DoA) acquisition and suppressive jamming elimination based on a Dynamic Agile Reconfigurable Intelligent Surface Antenna Array (DARISAA). DARISAA is a novel type of reconfigurable antenna array composed of many metamaterial elements, which can dynamically adapt its phase responses, enabling suppressive jamming cancellation at the antenna-end. Accurate DoA estimation schemes for jamming and desired signals are proposed based on a subspace approach by utilizing the dynamic agility of the DARISAA. By leveraging the spatial-domain preprocessing capability of DARISAA and the digital beamforming of the multi-channel receiver, an analog (antenna)-digital hybrid anti-jamming scheme is developed to effectively suppress high-power jamming and significantly improve the signal-to-interference-plus-noise ratio (SINR) even with low-resolution ADCs. Simulation results demonstrate that the proposed comprehensive anti-jamming scheme achieves significant SINR enhancement, highlighting the advantages of DARISAA-enabled spatial front-end processing for jamming suppression.
Read the full abstract
Physics and Society
Unveiling healthcare-access inequality in Ghana using a multiscale approach
Achieving universal health coverage, as set out in Sustainable Development Goal 3.8, requires closing persistent geographic and socioeconomic gaps in healthcare access, especially in under-resourced settings across Africa. Healthcare-access inequality is shaped not only by poor local access, but also by limited connectivity to city- and regional-level services and opportunities. Conventional accessibility analysis can identify where poor access occurs, but not whether poorly served places form structurally disconnected pockets across scales. This paper therefore builds on and extends the percolation divergence tree framework to develop a connectivity-based multiscale approach for examining healthcare-access inequality in Ghana. It combines street-level accessibility mapping with the hierarchical structure of the road network to identify the scales at which inequality coincides with connectivity breaks. The results first show substantial inequality: around one quarter of the population lives more than 5 km from the nearest healthcare facility. Multiscale analysis further reveals distinct structural forms of poor access. In relatively well-connected, monocentric regions, local poor-access pockets emerge around metropolitan fringes despite overall regional advantage. In less well-connected, polycentric regions, poor access extends across larger subsystems, with local pockets nested within broader poorly served areas. These findings show that healthcare-access inequality reflects both local conditions and the hierarchical connectivity of the wider spatial system, which may also constrain marginalised communities' access to other key resources and services. The framework can inform targeted local interventions and policy coordination across local and regional scales.
Read the full abstract
Logic
A complete representation theorem for nullnorms on bounded trellises
We establish necessary and sufficient conditions under which a binary operation on a bounded trellis is a proper nullnorm. The representation combines a t-conorm on the lower interval, a t-norm on the upper interval, two order-preserving maps, and a commutative, increasing function on $I_a^3\times I_a^3$, where $I_a^3$ consists of the elements incomparable with the absorbing element $a$ that neither reach $a$ nor are reachable from $a$. Unlike earlier range-restricted constructions, this function may take values anywhere in the trellis. To preserve associativity for such unrestricted values, we introduce a mixed interaction function that evaluates every pair with at least one component in $I_a^3$. We also derive the specializations in which this region is empty or consists of a single element, including an exact description of the admissible value in the singleton case. A five-element lattice example shows that the mixed associativity condition is independent of the remaining hypotheses, while a fourteen-element nontransitive trellis example demonstrates the necessity of allowing the unrestricted range. Finally, the principal range-restricted subclasses and the bounded-lattice case are recovered as specializations of the general representation.
Read the full abstract
High Energy Astrophysical Phenomena
First Light of Neutron Star Mergers: Off-axis Cocoon Cooling X-ray Emission from Short Gamma-Ray Burst Jets
Neutron star mergers (NSMs) are confirmed gravitational wave sources. Identifying an early electromagnetic counterpart for these events is therefore crucial for rapid localization and multimessenger follow-up. However, the associated Gamma-Ray Bursts (GRBs) are highly collimated and are therefore easily missed by off-axis observers. An early, less beamed counterpart is essential for identifying the majority of mergers. In this Letter, we investigate the cooling emission from jet-driven cocoons produced by short gamma-ray burst jets propagating through merger ejecta. We perform hydrodynamic simulations and radiative post-processing to calculate the early X-ray emission over a wide range of viewing angles. We find that the mildly relativistic cocoon produces bright soft X-ray transients for off-axis observers, with luminosities of $10^{46-48}{\rm erg\,s^{-1}}$ and durations of a few to ten seconds. The X-ray spectra are quasi-thermal with characteristic temperatures of $0.1$–$1\,{\rm keV}$. For observers at $\theta_{\rm v}=10^{\circ}$, the cocoon emission is detectable by Einstein Probe (EP) out to $z\simeq 0.3$. For nearby events like GW170817, it remains detectable up to $\theta_{\rm v}\simeq 45^{\circ}$. The predicted detection rate for EP is $0.5\,{\rm yr^{-1}}$ in the canonical model. In future multimessenger campaigns, rapid UV/optical/IR follow-up of such X-ray triggers can subsequently identify the associated kilonova and jet afterglow emission, which will confirm the origin.
Read the full abstract
Distributed Computing
Physically Partitioned KVCache Format for CPU–GPU Load Balancing in MoE Inference
Single-GPU long-context inference with Mixture-of-Experts (MoE) models requires spilling the key-value cache (KVCache) to CPU memory. The spilled KV serves two complementary purposes—transferring to the GPU for attention computation, or computing in-place on the CPU—which demand opposing physical states. The optimal split between them varies with workload, yet existing KVCache abstractions offer only storage semantics over a monolithic object of a single physical state, and cannot express dynamic load balancing. We propose InplaceKVCache, the first KVCache abstraction whose format fixes each byte's physical residency at write time, so that the CPU–GPU load balance can be adjusted without moving data after placement. It realizes this as a four-region layout along two dimensions—device affinity and access pattern—turning load balancing into pure scheduling. Built on this abstraction, WriteScope splits CPU–GPU shares along the sequence dimension, and a portable roofline performance model determines the optimal CPU share as sequence length evolves, with online feedback tracking CPU cost drift. On three MoE models (DeepSeek-V2-Lite, Qwen3-30B-A3B, Mixtral-8$\times$7B) with a 32~GB VRAM budget, WriteScope supports end-to-end inference at the 1M-token aggregate scale. In the long-context regime ($\ge$8K), it achieves geometric-mean speedups of $1.5\times$–$2.5\times$ on A100 and $1.4\times$–$1.7\times$ on V100 over four reproduced baselines, while vLLM, SGLang, and KTransformers fail even with a doubled KV budget. A DeepSeek-V4-Flash case study validates composition with native sparse attention.
Read the full abstract
Statistical Methodology
Bootstrap-Calibrated Spectral Divergence Tests for Online Detection of Covariance Matrix Changes
A covariance matrix rarely distorts in a single direction: a shift can expand all variances simultaneously, shift variance along one or two principal directions, displace spectral mass without changing marginal means, or rotate the dependence structure. Default mean-shift detectors are blind to these effects, and no existing online procedure calibrates a family of spectral deviation tests with provable false alarm control across both time and tracking window selection; this paper fills that gap. We consider four spectral deviations: $D_{KL}(P_{1}\|P_{0})$, $D_{KL}(P_{0}\|P_{1})$, Jeffreys, and Bhattacharyya, each evaluated on the eigenvalues of the empirical relative covariance operator $\widehat{\Sigma}_{0}^{-1/2}\widehat{\Sigma}_{t}\widehat{\Sigma}_{0}^{-1/2}$. Critical values come from a conditional parametric bootstrap that accounts for estimation uncertainty in both the past and tracking windows, an aspect asymptotic approaches typically overlook. The procedure controls the false alarm rate family-by-family over a predefined monitoring period and a range of candidate window sizes; when a single operational window is needed, a power-based criterion selects it. We prove consistency under constant alternatives via local spectral expansions with second-order sensitivity near the null. Simulations are conservative under the null and show detection power depends largely on spectral shape rather than magnitude: $D_{KL}(P_{1}\|P_{0})$ excels under global inflation, while Jeffreys and Bhattacharyya cover a broader range of alternatives. We illustrate the approach on three financial applications: European stock indices, Fama-French sector portfolios, and large-cap technology stocks.
Read the full abstract
Differential Geometry
The Cartan-Hadamard isoperimetric inequality under controlled variation of curvature
We prove the Cartan-Hadamard conjecture (the sharp Euclidean isoperimetric inequality) in every dimension, assuming that the sectional curvature is pinched relative to a reference radius of curvature. The reference radius of curvature may vary, and the condition includes examples where the radius of curvature blows up, and the manifold has both unbounded (negative) and asymptotically vanishing sectional curvature. The key step in the proof is to control the entire convex hull of a quotient minimiser at its mean-curvature scale and compare the metric distortion with the surplus in the normalised hyperbolic isoperimetric profile at that same scale.
Read the full abstract
Machine Learning
Should All Noises Be Treated Equally: Impact of Input Noise Variability on Neural Network Robustness
Geophysical data collected from active field sites are often contaminated by complex and heterogeneous noise, obscuring weak seismic events, and complicating automated interpretation. Although deep learning offers promising solutions for seismic processing, its performance is highly sensitive to the nature of training noise, especially under out-of-distribution (OOD) conditions. This study investigates the influence of noise parameters, such as type, scale, and complexity on the performance, generalization and robustness of neural networks in two geophysical tasks: first break picking and denoising. We simulate seismic-while-drilling data and apply controlled input source noise augmentation using stochastic generators to vary the noise characteristics. Different neural networks are trained on fixed noise types and scales, then evaluated across both seen and unseen noise scenarios. We incrementally increase the complexity of the noise by introducing compound noise mixtures and assess the performance of the model under increasingly challenging OOD conditions. This yields a robustness matrix that captures the generalizability of each model relative to its training configuration. Results indicate that larger noise scales boost generalization, and that effective alignment between noise type, task complexity, and architecture is key for maximizing generalization gains. In addition, training with compound noises mitigate weaknesses associated with single-noise training, acting as an additional implicit regularizer to improving robustness. These findings highlight key factors influencing model resilience in noisy geophysical environments and offer guidance for developing deep learning models that generalize effectively across diverse and unpredictable noise conditions.
Read the full abstract
Statistical Methodology
Identification of causal policy effects for spatial exposure fields
Many interventions act on an entire spatial exposure field. We study when their induced laws satisfy policy positivity and what its failure implies for causal evaluation through weighting and structural modelling. We show that, for post-transformation fields that are Gaussian, the only admissible translation that satisfy positivity lie on the Cameron-Martin space of the observed exposure field. Under fine-scale variation and regularity assumptions, these are the only admissible non-decreasing pointwise policies on that scale. Common rules, including binding caps and proportional reductions on the Gaussian scale can therefore violate positivity. The Cameron-Martin norm controls both weight variability and the amplification of disagreement between candidate outcome models into disagreement between policy estimates. Without positivity, bounded continuous response models can make arbitrarily similar observational predictions while their policy estimates remain separated. We also give a policy-specific criterion for structural identification conditional on a realised exposure state. Finally, when a policy violating positivity admits weights in every finite representation, their second moments diverge under nested refinement that recovers the full field. These results motivate assessing identification and sensitivity for the specified policy effect across plausible structural models and under refinement of finite representations.
Read the full abstract
Analysis of PDEs
Billiards, Refraction, and Blow-Up for Kirchhoff Equations
We start from a simple problem in geometric optics. A particle moves between two homothetic ellipses, with refraction at the inner interface and reflection at the outer one. In a suitable nearly circular regime, the small geometric anisotropy of the ellipses is amplified by the strong refraction, and the corresponding return map develops a transverse heteroclinic connection between the two axial motions. We turn this geometric mechanism into a construction for a two-mode Kirchhoff system while keeping the standard quadratic elastic variable throughout. The singular optical model is first reduced to an explicit kick–drift map, for which the heteroclinic connection is obtained by a contraction argument. The same argument also gives exponential convergence along the two tails and transversality. We then show that the connection persists for the genuine elliptic billiard and, subsequently, through a smooth regularization of the reflecting and refracting interfaces. Finally, a small positive background is added to the Kirchhoff coefficient, making it uniformly positive without destroying the hyperbolicity of the axial modes or the transverse connection. This produces a smooth and uniformly positive Kirchhoff coefficient admitting a transverse heteroclinic orbit between two simple modes. Combined with the road-map theorem developed in our previous work, the construction yields a finite-time blow-up example for a forced abstract Kirchhoff equation with a regular forcing term.
Read the full abstract
Numerical Analysis
Hankel-Koopman Finite-Horizon Energy Decomposition of Coupled Experimental Data: A Three-Phase Data-Driven Twin Forecasting Framework
Finite-horizon prediction is challenging when observations are limited and a quantity of interest depends on several coupled nonlinear measurement channels. This paper develops a three-phase data-driven twin framework that integrates finite-horizon Hankel–Koopman energy decomposition, inverse-calibrated reconstruction using nonlinear autoregressive models with exogenous inputs, and forecast composition. Channel dynamics are identified and simulated in Hankel coefficient space, with candidate models assessed after reconstruction in physical measurement space. The resulting reduced-order model of the channel dynamics drives recursive prediction of the quantity of interest. Three theorems support the framework, establishing the energy decomposition and modal orthogonality; existence, uniqueness, and inverse stability of the auxiliary continuous inverse-calibration solution; and finite-horizon stability of the composed forecast for a fixed selected model under stated assumptions. A retrospective solar-power case study yields high correlations between predicted and reference trajectories across the considered forecast horizons, demonstrating reproduction of a constructed channels-dependent quantity of interest.
Read the full abstract
Artificial Intelligence
When does a scaling result justify a different allocation? A critical review of resource-allocation evidence for AI systems
AI scaling studies increasingly evaluate systems that combine a pretrained model with retrieval, search, verification, tools, and interaction. Yet a higher score under a larger budget does not by itself show where additional resources are best spent. This critical integrative review asks when a reported scaling result supports a resource-allocation decision. It compares evidence across pretraining, test-time computation, retrieval, and agent evaluation, distinguishing the performance of a tested procedure from the best performance achievable under a resource limit. The synthesis shows that three mismatches recur across this evidence: success counted before an answer is chosen, information a deployed system will not have, and costs left out of the comparison. A capability surface expresses performance as a function of budgets, mechanisms, and available information. Worked analytical examples show how the evaluation metric, deployment volume, selection rule, and stopping policy can alter an allocation conclusion. A resource envelope provides a structured record of the task, development and run-time resources, information access, and procedure behind a reported score. Its application to a published comparison illustrates which conclusions the evidence supports and which deployment questions remain unresolved. The resulting framework specifies the comparisons needed to choose among feasible systems and motivates experiments on the transfer of allocation rules across tasks and operating conditions. It does not propose a universal scaling law or infer general intelligence from benchmark gains.
Read the full abstract
Materials Science
Contribution of photogenerated charge carriers to photothermal effect in optically opaque semiconductor samples
As a consequence of laser radiation absorption, the photothermal response of semiconductors is governed by both the thermalization of the crystal lattice and the dynamics of photogenerated charge carriers. This carrier-related contribution leaves a characteristic signature in the temperature field through carrier diffusion and recombination, making the photothermal response sensitive to electronic transport and recombination properties. This sensitivity provides the basis for using photothermal methods to characterize the electronic properties of semiconductors. In this paper, we develop a theoretical model of the photothermal response of moderately doped opaque semiconductors under harmonically modulated laser excitation, accounting for the generation, diffusion, and recombination of minority charge carriers. We analyze how carrier lifetime, diffusion length, and surface recombination velocity affect the resulting temperature field and, consequently, the photothermal signal. In particular, we show that the thermal contribution associated with carrier recombination remains nonzero even when the surface recombination rate vanishes and for short carrier lifetimes, demonstrating that the signature of photogenerated carriers may persist under conditions where their contribution might otherwise be expected to become negligible. The results provide guidance for both the design and interpretation of photothermal and photoacoustic experiments, including the choice of modulation frequency and detection geometry, to enhance the sensitivity to carrier-related thermal signatures and improve the determination of electronic properties of semiconductor materials and devices.
Read the full abstract
Machine Learning
EdgeHAR: An Edge-Native Compact Sensor Foundation Model for Human Activity Recognition
Sensor-based human activity recognition (HAR) is fundamental to ubiquitous and wearable computing, yet existing foundation models are largely designed for cloud-scale deployment and struggle with real-world sensing shifts, including unseen users, devices, sampling rates, and sensor placements. We present EdgeHAR, an edge-native compact sensor foundation model designed for wearable intelligence. Unlike conventional models that entangle activity knowledge with acquisition variations, EdgeHAR learns transferable representations by factorizing sensor signals into three latent codes: an (i)Activity-Semantic Code capturing reusable activity knowledge, a (ii)Motion-Dynamics Code modeling temporal patterns, and an (iii)Acquisition-Context Code representing sensor-specific variations. This disentangled design enables efficient adaptation to new users, devices, placements, and activity classes with limited target-domain data. By incorporating lightweight adaptation modules, EdgeHAR achieves foundation-model-level transferability while satisfying edge constraints in computation, memory, latency, and privacy. Experiments across heterogeneous HAR datasets demonstrate that EdgeHAR maintains competitive recognition performance under distribution shifts with substantially reduced deployment cost. EdgeHAR establishes a practical paradigm for compact, edge-first foundation models for ubiquitous sensing systems.
Read the full abstract
High Energy Physics (Theory)
M2 brane in AdS$_4\times S^7/\mathbb Z_k$: 2-loop correction to $\frac{1}{2}$-BPS Wilson loop vs localization in ABJM theory
We study the quantization of an M2 brane wrapping $\mathrm{AdS}_2\times S^1$ inside $\mathrm{AdS}_4\times S^7/\mathbb{Z}_k$, dual to the $\frac{1}{2}$-BPS circular Wilson loop in ABJM theory. The classical action and the one-loop contribution were previously shown to reproduce the gauge-theory localization result. Here we extend the analysis to the two-loop order. In an adapted static and $\kappa$-symmetry gauge the world-volume action contains no cubic interaction vertices, so that the two-loop correction is determined entirely by quartic interactions and is manifestly free of logarithmic UV divergences. We find that the bosonic and fermionic contributions combine into a perfect square $ {\mathscr X}^2$ where ${\mathscr X}$ is a linear combination of the bosonic and fermionic 3d Green's functions and their derivatives. The evaluation of $\mathscr X$ requires fixing the $\mathrm{AdS}_2$ boundary quantization of the massless fermionic modes. With the boundary conditions selected by the preserved supersymmetry, $\mathscr X$ vanishes and hence the complete two-loop correction is found to be zero. This result would not be consistent with identifying the M2-brane loop expansion directly with the large-$N$ expansion of the canonical, fixed-$(N,k)$ localization prediction. Instead, it agrees precisely with the recent proposal that the semiclassical M2-brane partition function should be matched to the gauge-theory grand-canonical ensemble, in which the Wilson loop expectation value is perturbatively one-loop exact. The vanishing of the two-loop correction thus provides a non-trivial test of this grand-canonical identification.
Read the full abstract
Numerical Analysis
Batched Wavefront Sweeps for Linear Transport Uncertainty Quantification
We develop a graph-compatible batched wavefront formulation for upwind discontinuous Galerkin discretisations of linear transport. For sweepable discretisations, the cell equations form a block lower-triangular system under a topological ordering of the directed upwind dependency graph. This causal structure can be shared across families of fixed-source problems whose local operators, sources and inflow data vary. We exploit this observation by grouping channels with a common dependency graph into graph-compatible sweep classes and carrying samples, right-hand sides, energy groups and sign-compatible angular ordinates as tensor dimensions within a common wavefront schedule. We prove that the resulting batched wavefront algorithm is algebraically equivalent to independent classical DG sweeps for every channel in a class, while exposing parallelism simultaneously across wavefront cells and channel dimensions. We realise the formulation as a GPU tensor program based on batched cell-local solves and quantify its work, depth and storage requirements, including the throughput–memory tradeoff introduced by sample microbatching. We further characterise the discrete parameter-to-observable map on fixed face-sign branches and show that reverse-mode differentiation through the wavefront algorithm reproduces the corresponding discrete adjoint action. Numerical experiments verify the expected DG convergence and adjoint consistency, demonstrate substantial GPU execution gains from sample batching, and illustrate the method in a material-shadowing uncertainty-quantification problem and a coupled multigroup C5G7/KAIST-inspired power-iteration workload.
Read the full abstract
Computer Vision
MCIQA-2K: A Multi-Dimensional Dataset and No-Reference Quality Assessment Benchmark for Colorized Images
Image colorization is an inherently ill-posed task, since a single grayscale image may correspond to multiple plausible colorized results. Consequently, conventional full-reference image quality assessment (IQA) metrics fail to accurately reflect human perceptual preferences for colorized images. In this paper, we present MCIQA-2K, a large-scale multi-dimensional benchmark specifically designed for no-reference quality assessment of colorized images. We construct a dataset containing 2,000 colorized images generated by five representative colorization models, together with human annotations across three perceptual dimensions: color smearing, semantic color misalignment, and global naturalness. Building upon the proposed benchmark, we further introduce MCIQA, a dedicated multi-branch NR-IQA framework for colorized images. Extensive experiments demonstrate that MCIQA significantly outperforms existing full-reference and no-reference IQA methods on the proposed benchmark, while also exhibiting competitive generalization capability on several widely-used IQA datasets. The dataset and code are publicly available at https://github.com/ARBEZ-ZEBRA/MCIQA.
Read the full abstract
High Energy Astrophysical Phenomena
NICER neutron stars with dark energy and dark matter: effects on the inferred equation of state
[Abridged] In this work, we compare neutron stars where dark energy or dark matter are present in the core, contrasting them against purely baryonic stars with the same central energy density. From this comparison, we find that the dark energy models are more constrained by neutron star observations than the dark matter models, suggesting that neutron stars can place strong constraints on dark energy parameters. We consider four configurations: purely baryonic neutron stars described by a piecewise polytropic (PP) equation of state (EoS), baryonic stars admixed with bosonic or fermionic asymmetric dark matter (ADM) cores, and stars with a dark energy core and baryonic shell, where the dark energy is described by the modified Chaplygin dark fluid (MCDF). Using the masses and radii of PSR J0740$+$6620, PSR J0437$-$4715, and PSR J0030$+$0451, we employ Bayesian inference to investigate how accounting for bosonic/fermionic ADM and MCDF cores affects the inferred neutron star properties. We find that MCDF cores, more than PP and ADM admixed models, substantially broaden neutron star mass-radius and pressure-energy density posterior distributions. Moreover, the MCDF EoS parameters can be tightly constrained. While bosonic and fermionic ADM cores decrease the inferred maximum mass, their posteriors strongly coincide with those of the PP model for ADM mass-fractions $F_\chi \leq 5\%$. These results demonstrate that current mass-radius observations cannot rule out MCDF cores, but they can constrain the MCDF parameter space to narrow regions. ADM admixed neutron stars, however, remain observationally identical to purely baryonic stars for all explored mass-fractions, even with improved mass-radius uncertainties. This suggests that mass-radius measurements alone are insufficient to detect ADM core inside neutron stars and that independent astrophysical or particle physics probes will be required.
Read the full abstract
Information Retrieval
VARG: Value-Aware and Ranking-Aligned Generative Retrieval for Dynamic E-commerce Search
Integrating recall and pre-ranking in e-commerce search requires candidate generation to account for relevance, personalization, and business value before final ranking. To this end, we present VARG, a generative retrieval system for Tmall App search that directly admits generated item candidates to the existing final ranker. VARG-ID constructs semantic prefixes using RQ-VAE, enhances search relevance through bidirectional query-item contrastive learning, and combines these prefixes with a value-ordered third token to provide fine-grained item addresses and a business-value prior. Three-stage supervised fine-tuning progressively learns item-to-identifier mappings, query-semantic retrieval, and personalized retrieval. Personalized model training combines value-aware and hierarchy-aligned supervision with expanded user context, and uses local ordinal supervision (LO-SFT) to learn the local within-cluster ordering encoded by the third token. Prefix-GRPO combines gated rewards based on output legality, user behavior, ranker advantage, and search relevance with prefix-aware token weighting to align candidate generation with business value and ranking objectives. Coordinated daily product and model updates preserve existing item addresses while incorporating new products and behavioral feedback. Offline experiments on tens of millions of products validate identifier stability and demonstrate gains in retrieval quality and head-level value recall from SFT strategies and Prefix-GRPO over their respective baselines. In a 14-day online A/B test covering 20% of search traffic, VARG directly admits generated candidates to the final ranker and improves GMV by 1.45%, per-user IPV by 0.22%, and PCTR by 0.31%. Online shopping-guide query evaluations further show that VARG maintains competitive relevance with a smaller candidate quota.
Read the full abstract
Number Theory
Averaged dynamical generalizations of Chowla's conjecture
In 1965, Chowla conjectured that the signs of the Liouville function become asymptotically uncorrelated at any fixed collection of distinct shifts. In 2015, Matomäki, Radziwiłłand Tao proved an averaged form of Chowla's conjecture. In 2022, Lichtman proved a variant of this conjecture over primes on average. In the same year, Lichtman and Teräväinen proved the Hardy-Littlewood-Chowla conjecture on average. In this article, motivated by the recent work of Bergelson and Richter on the dynamical generalizations of the prime number theorem, we will first establish the analogous dynamical generalizations of these three results related to Chowla's conjecture. In the proofs, we follow an approach of Qi and Zheng who established a variant of Bergelson and Richter's theorem over irreducible binary cubic forms. In 2016, Tao proved that the two-point logarithmic Chowla's conjecture holds. Recently, Charamaras and Richter generalized Tao's theorem to bounded arithmetic functions and proposed a conjecture that generalizes Chowla's conjecture to bounded multi-variable arithmetic functions. Building on their work, we will also prove a dynamical generalization of Tao's theorem and some variants for the composition of the sum-of-digits function with the prime Omega function.
Read the full abstract
Quantitative Methods
MIRAGE: Measuring Interpolation and Redundancy in Affinity GEneralization
Deep learning now underpins structure-based drug design, from complex and affinity prediction to ligand ranking and pose generation. Recent co-folding models reportedly approach free-energy-perturbation accuracy at far lower cost. Yet standard evaluation, a single held-out correlation or pooled pose-success rate, cannot separate transferable binding principles from repeated exposure to related protein families in public databases, and practical success depends on genuinely novel targets. We introduce MIRAGE (Measuring Interpolation and Redundancy in Affinity GEneralization), a plug-in benchmark treating historical public family support (through 2019) as an explicit variable, applying a family-support axis to affinity and pose prediction via matched strata, family-disjoint controls, ligand-only baselines, and temporal evaluation. Co-folder affinity accuracy rises sharply with family support, while shallow controls that cannot exploit the test family stay flat, large for co-folders and near zero for every family-disjoint or trivial control. For Nesso-1 it survives covariate, conditioning, balancing, and clustering checks; Boltz-2's endpoint is limited by coverage. It localizes to family support rather than ligand chemistry, approaching a level from family identity alone. Rankings reverse on novel families, where a family-disjoint random forest leads both co-folders, significantly vs Nesso-1. On one external low-support target, neither co-folder beats molecular weight, corroborative rather than population-level evidence. gnina shows significant support dependence in rescoring whereas smina does not; MSA-free pose engines show larger gaps than smina redocking. This redundancy-driven inflation differs from conventional leakage. We propose reporting performance across family support plus excess over a support-insensitive baseline, and release MIRAGE as an installable benchmark and dataset.
Read the full abstract
Formal Languages and Automata
Compressed Subsequence Checking is PSPACE-complete
It is shown that the (scattered) subsequence problem for two words represented by straight-line programs is $\mathsf{PSPACE}$-complete, even over a binary alphabet. The lower bound is obtained by a polynomial-time reduction from quantified subset sum. If the two words are given explicitly, i.e., uncompressed, then the subsequence problem is hard for uniform $\mathsf{NC}^1$ and in deterministic logspace.
Read the full abstract
Classical Analysis and ODEs
The `1/9'-problem for the $\varphi$-functions
The functions $\varphi_\ell(z)=(e^z-s_\ell(z))/z^\ell$, with $s_\ell$ the Taylor polynomial of $e^z$ of degree $\ell-1$, are what exponential integrators evaluate, and they evaluate them through rational approximations on the negative real axis $\mathbb{R}_-$. Schmelzer and Trefethen (2007/08) conjectured that the best such approximations converge at the rate of the classical `1/9'-problem: $E_n(\varphi_\ell)^{1/n}\to H$, Halphen's constant, for every $\ell$. We prove it. The theorem behind it gives the same rate for $f=u_0+u_1\exp$ with $u_0$ rational and $u_1\not\equiv0$ meromorphic with finitely many poles and $\log|u_1(z)|=o(|z|)$, the lower bound required only away from $\mathbb{R}_-$, whenever $f$ has no pole on $\mathbb{R}_-$ (Theorem~4.4). For rational $u_1$ this is Theorem 1 of Stahl and Schmelzer (2009), announced with its proof deferred to a manuscript that never appeared; the hypotheses admit, for instance, $u_1(z)=\sinh(\pi\sqrt z)/(\pi\sqrt z)$, whose zeros are infinite in number. The proof is a reading of Gonchar and Rakhmanov (1989) rather than an extension of them. Their theorem concerns a sequence given by a Cauchy-type integral with a varying weight $\Phi_n$, and a contour identity of Schmelzer and Trefethen presents $u_1\exp$, once its principal parts are removed, in exactly that form, with $\Phi_n(t)=u_1(-nt)e^{-nt}$. The factor $u_1(-nt)$ grows subexponentially in $n$, and their hypothesis on $\Phi_n$ divides by $2n$, so it drops out of the external field: field, extremal arc, $S$-property and constant are all theirs, unchanged. For $\varphi_\ell$ the factor is $t^{-\ell}$, whose pole sits at the origin, a point of the approximation set and hence at positive distance from every admissible contour.
Read the full abstract
Astrophysics of Galaxies
The Connection between Halo Assembly History and the Stellar Distribution I: The Central Stellar Mass - Richness Plane
Recent observations suggest intrinsic links between the stellar components of massive halos and their assembly histories. Using massive halos with $\log[M_{\rm halo}/M_\odot] \geq 13.0$ from the IllustrisTNG-300 simulation, we explore differences in secondary halo properties between samples selected by central galaxy stellar mass and satellite richness. For group-like masses, we find that halos selected by the outskirts stellar mass of their central galaxies between 50 and 100 kpc ($M_{\star,[50,100]}$) have similar halo mass distributions to those selected by intrinsic satellite richness ($\lambda_{10, R_{200}}$), but differ significantly in assembly history: $M_{\star,[50,100]}$-selected halos are more concentrated and formed earlier. At higher masses ($\sim10^{14}~M_\odot$), intrinsic richness becomes the better halo-mass proxy, but only when projection effects are ignored. We find that the difference-set test selects comparable halo-mass samples most effectively when the two proxies have similar scatter. Projection effects and baryonic physics do not alter the main trends but remain important caveats for observational applications. Our findings show that the central stellar mass-richness plane is a powerful tool for selecting halos with particular accretion histories, offering new insights into the galaxy-halo connection in the era of deep imaging and precision lensing surveys such as Euclid and LSST.
Read the full abstract
General Economics
Equilibrium Transition and Cartel Formation: A Structural Analysis of Chile's Pharmacy Cartel
This paper studies how Chile's three largest pharmacy chains moved from the price war to collusion, using court-record daily prices and a structural model. After a court-ordered advertisement ban ended the comparison campaign, the estimated gain from loss-leader pricing fell sharply, weakening the incentive to continue the price war. The chains then used an upstream supplier as intermediary to verify a collusive price-leadership procedure. Once verified, they applied it to restore margins on former loss leaders. They then raised prices further to extract rents. These increases spread more slowly than margin restoration. The Adaptive Confidence specification allows a subjective belief about follower participation to update. It more closely matches the cumulative spread of rent extraction than the Unit Confidence specification. Unit Confidence imposes the rational-belief benchmark by fixing the leader's weight at one.