Peer Revenue Publish or profit

Papers

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

All of arXiv

Friday 4 September

Algebraic Geometry

Legendrian families of lines on nilpotent orbit closures

Let $\mathfrak{g}$ be a complex semisimple Lie algebra, and $\overline{Z}$ a nilpotent orbit closure in the projectivization $\mathbb{P}(\mathfrak{g})$. We investigate the space of tangent directions of lines on $\overline{Z}$ passing through a general point $z$, denoted by $F(\overline{Z},\,z)$. We first prove that every irreducible component of $F(\overline{Z},\,z)$ is an integral subvariety of the contact hyperplane in the projectivized tangent space. Next, we study Legendrian components of $F(\overline{Z},\,z)$ in two cases: the case where $\overline{Z}$ admits a Springer resolution; the case where $\overline{Z}$ is square-zero in a projectivized simple Lie algebra of classical type. As an application, two corollaries are presented: a characterization of Richardson orbits among square-zero orbits in terms of $F(\overline{Z},\,z)$; a description of $F(\overline{Z},\,z)$ for $\overline{Z}$ arising from stratified Mukai flops associated to irreducible Hermitian symmetric spaces.
Read the full abstract

Formal Languages and Automata

On the equivalence between generating functions computed by memory transducers and enumerating functions produced by indexed grammars

We consider the sequences of natural integers that can be computed by a deterministic transducer, with input in a structure A, output in N. and with memory the set of stacks of stacks of A. We show that these sequences are, exactly, the counting sequences of formal languages generated by unambiguous context-free indexed grammars (equivalently, the counting sequences of derivation trees of arbitrary context-free indexed grammars), with indexes in A. This general theorem applies, notably, to the set of natural integers endowed with the operation -1 and the non-zero predicate, showing that the polynomial recurrences count exactly the index-languages (where the parameter used for counting is the index itself).
Read the full abstract

Optics

Fabrication-Aware Design of a Hybrid Metasurface-Bragg Mirror for Low-Noise Precision Optics

High-reflectivity coatings for precision interferometry must simultaneously minimize optical losses and thermally driven displacement noise. Dielectric Bragg mirrors provide robust high reflectance but rely on thick multilayer coatings, whereas metasurface mirrors provide high reflectance and low noise but are sensitive to fabrication-induced deviations. A fabrication-aware photonic design framework is introduced that combines a resonant single-layer metasurface, an etch-stop layer, an antiresonant spacer, and a reduced Bragg reflector. Fabrication effects, including geometric tolerances and line-edge roughness, are explicitly considered in the optical design. Full-wave electromagnetic simulations are combined with a truncated-Gaussian Monte Carlo analysis to determine the performance distribution under the assumed fabrication conditions. The ideal metasurface design achieves a modeled non-reflected power of $1-R<10^{-5}$. After roughness-aware reoptimization, the modeled non-reflected power remains below $1.04\cdot10^{-4}$ at $80\%$ fabrication yield. For the cryogenic ETpathfinder gravitational-wave testbed, two Bragg layer pairs reduce the non-reflected power of the complete stack to approximately $6.6\,\mathrm{ppm}$. The estimated thermal displacement-noise amplitude spectral density is $1.97\cdot10^{-20}\,\mathrm{m/\sqrt{Hz}}$ at $100\,\mathrm{Hz}$. The architecture connects fabrication robustness, optical performance, and thermal-noise reduction within a single design framework.
Read the full abstract

All authors Mika Gaedtke, Christian Kranhold, Markus Walther, Falk Eilenberger, Stefanie Kroker, Thomas Siefke

Materials Science

Local coordination, structural softening, and polarization-switching energetics in Sc-alloyed GaN

Sc-alloyed wurtzite nitrides exhibit strongly tunable electromechanical and ferroelectric properties, yet the relationship between their lattice evolution and local bonding environment remains insufficiently established, particularly in Sc$_x$Ga$_{1-x}$N. Here, we investigate the structural and local bonding evolution of Sc$_x$Ga$_{1-x}$N across the dilute-to-intermediate composition range by combining X-ray diffraction, Sc K-edge X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS) and first-principles calculations. Sc incorporation produces an anisotropic lattice expansion and a progressive reduction in the c/a ratio. XANES and EXAFS reveal a concurrent modification of the local environment around Sc, with the effective Sc-N coordination number increasing from 4.1(4) to 4.5(2) and the average Sc-N bond length increasing from 2.045(7) to 2.081(8)Å over x=0.06-0.26. The local response is accompanied by a reduction in the Sc K-edge pre-edge intensity, consistent with a gradual reduction of the local tetrahedral asymmetry. First-principles calculations show that these structural changes are associated with site-selective distortions around Sc and progressive flattening of the structural energy landscape. The calculated intrinsic polarization-switching barrier decreases from 24.2 to 19.0 meV/Å$^3$ with increasing Sc content, while the calculated piezoelectric stress coefficient (e$_{33}$) increases from 0.82 to 1.66~C/m$^2$ and elastic constant (C$_{33}$) decreases from 380 to 227~GPa, resulting in an increase of piezoelectric strain coefficient (d$_{33}$) from 2.99 to 12.46~pC/N. These results show that Sc incorporation progressively modifies the local coordination environment and structural energetics of ScGaN while the long-range wurtzite structure remains preserved over the investigated composition range.
Read the full abstract

All authors Shailesh Kalal, Gueorgui Kostov Gueorguiev, Martin Magnuson, Edward Ferraz de Almeida Junior, Rohini Sanikop, Sagar Jathar, Per Sandström, Ray-Hua Horng, Jens Birch, Per Eklund, Ching-Lien Hsiao

Computation and Language

BIT.UA at BioASQ 14B: Modular Retrieval with pg_textsearch and Qdrant, and Agent-Based Answer Generation

This paper describes the participation of the BIT.UA team from the University of Aveiro in the 14th edition of the BioASQ Task B challenge on biomedical question answering. Building on our previous submissions, we introduced a substantially refactored and modular codebase, and made significant changes to both the retrieval and generation components of the pipeline. For Phase~A document retrieval, we replaced the PyTerrier PISA index with PostgreSQL-based pg\_textsearch for BM25 retrieval and adopted Qdrant for dense embedding indexing, enabling more efficient storage and GPU-accelerated similarity search. We explored HyDE-based query expansion alongside a Context-1 retrieval strategy. A new reranker training pipeline was developed, incorporating dense retrieval for negative sampling. For Phases A+ and B answer generation, we introduced an LLM-as-a-judge framework and a novel agent quorum mechanism, where multiple agents with diverse prompts debate and iteratively converge on a consensus answer using adaptive document retention. We also participated in the snippets generation subtask for the first time. Our systems achieved competitive results across all batches, with Phase~A systems achieving MAP ranks of 5 (Batch~1,3). We discuss the impact of these architectural changes, lessons learned, and outline directions for future work including SPLADE and ColBERT integration. All code is openly available: https://github.com/bioinformatics-ua/BioASQ14b.
Read the full abstract

All authors André Ribeiro, Rúben Garrido, Alexander Christiansen, Richard A. A. Jonker, Sérgio Matos

High Energy Astrophysical Phenomena

The October 2022 flare in OJ 287 and the mass of its primary black hole

The bright blazar OJ~287 has demonstrated a sequence of flares, which are well explained by a quasi-Keplerian orbit model. The flares are associated with the impact of the secondary on the accretion disk of the primary. The orbit must precess in order to produce the correct sequence of flares, and from the precession rate we calculate the mass of the primary. This precession rate gives the mass of the primary $M_{BH} = (18.35\pm0.05) \times 10^9 M_{\odot}$. Two kinds of flares have been identified: direct flares from the impacts, and tidal flares arising from an increased accretion flow into the jet. The precession rate and the primary black hole mass may be independently determined from both sets of flares; the tidal flare of October 2022 was recommended for an intense campaign for this reason. This paper describes these observations over a wide spectral range. We show that the October 2022 flare fits the expectations for a tidal flare and thus supports the earlier determination of the mass of the binary black hole system in OJ 287. The mass of the primary may also be deduced from secondary indicators such as the correlation with the hydrogen line strength and the black hole mass. These studies require that the mass is above $M_{BH} \sim 10^{10} M_{\odot}$, but do not specify the value more exactly.
Read the full abstract

All authors Mauri J. Valtonen, Staszek Zola, Manpreet Singh, Andrei V. Berdyugin, Kari Nilsson, A. Gopakumar, Alok C. Gupta, Tapio Pursimo, Alexandr E. Volvach, Marek Drozdz, Waldemar Ogloza, Rene Hudec, Martin Jelínek, Jan Štrobl, Michal Zejmo, Stefano Ciprini, Daniel E. Reichart, Vladimir V. Kouprianov, Alberto Sadun, Markus Mugrauer, Katsura Matsumoto, Ryo Imazawa, Makoto Uemura, Lankeswar Dey

Quantum Physics

Learning unknown stabilizer codes using product measurements

Efficiently characterizing quantum error correcting codes is a key challenge on the path to fault-tolerant quantum computation. Stabilizer codes, a central class of such codes, are defined by a set of stabilizer generators. Here, we present an algorithm that uses random single-qubit measurements to learn the stabilizer generators of any stabilizer code from $N$ copies of stabilizer states in its codespace, requiring no prior knowledge of the code's structure. This also enables verification that a device implements its intended code. We derive a lower bound on $N$ needed to recover the stabilizer generators with high probability, together with a bound on the algorithm's overall probability of success. When applied to quantum low-density parity-check (qLDPC) codes, a leading candidate for practical fault-tolerant architectures, our approach requires a number of states that scales polylogarithmically with $n$, the number of qubits.
Read the full abstract

Optics

Fisher-information retention under local driving in non-Hermitian feed-forward chains

When local drive hardware is installed before a sensing task is revealed, directional non-Hermitian buildup creates a tradeoff between targeted preparation and spatial coverage. We quantify this tradeoff for stable linear scatterers with proper complex Gaussian output statistics. For parameters encoded in the internal generator, and at fixed coherent internal resource, the drive hardware enters the optimized mean-response Fisher information only through the projector onto the internally reachable subspace, while the preparation Gramian sets the source cost. In an on-resonance feed-forward chain, nested Green-function columns decompose into disjoint port intervals, and the site-local retention factors in each interval sum exactly to unity. Directionality concentrates this unit downstream, so the worst-site retention decays exponentially when one port serves many sites; for $m$ equal intervals, a single probe shared by all tasks has a worst-site retention lower by a further factor $m$ than task-specific reoptimization. Joint source-power and internal-resource constraints connect this fixed-resource law to source-limited preparation. A noisy RLC/VCCS model illustrates the retention law with stored electromagnetic energy, covariance Fisher information, and component disorder. The resulting bounds turn directional buildup into a port-layout criterion: source savings for a selected task are accompanied by a spatial coverage cost for tasks unknown at the hardware-design stage.
Read the full abstract

Machine Learning

Beyond Homoscedasticity: Decoupled Uncertainty Optimization for Deep Imbalanced Regression

Deep Imbalanced Regression (DIR) is pervasive in continuous prediction tasks across diverse modalities, such as age estimation, depth prediction, and protein mutation activity prediction, where label-scarce tail samples often carry higher practical value. However, most existing methods still learn deterministic point mappings under mean squared error or its simple variants, implicitly assuming a uniform uncertainty level across all samples and thereby overlooking the instance-wise heteroscedasticity that is widespread in long-tailed data. We further point out that even heteroscedastic negative log-likelihood suffers from a gradient coupling issue, which, under DIR scenarios, weakens the learning signal of hard tail samples and leads to optimization inertia as well as tail underfitting. To address this, we propose DUO, an uncertainty-aware long-tailed regression framework. Specifically, the proposed method models the regression target as a conditional Gaussian distribution to explicitly characterize instance-level predictive uncertainty, and transforms uncertainty into a dynamic enhancement signal for tail samples through decoupled mean-variance optimization. Furthermore, we design a distribution-guided contrastive learning mechanism that adaptively constructs positive and negative pairs based on the overlap between sample distributions, thereby alleviating feature looseness and cross-label semantic entanglement. Across visual and biological DIR benchmarks, DUO achieves the best few-shot bMAE and GM on IMDB-WIKI-DIR, AgeDB-DIR, and AAV2-DIR while remaining competitive on few-shot MAE.
Read the full abstract

All authors Juncheng Zhou, Jiaxi Lu, Weijing Zeng, Zhong Li, Hao Qi, Jingsong Cui

Econometrics

Identification and Estimation of Intergenerational Income Mobility Measures

Measuring the intergenerational transmission of lifetime economic status is complicated by researchers often only observing snapshots of income at specific ages. Consequently, standard practice estimates intergenerational mobility using income averages, introducing life-cycle bias that compromises reliability and comparability across studies, time, and place. I develop a missing data framework that exploits available income data and observable characteristics to eliminate life-cycle bias. This method combines nonparametric identification with Neyman-orthogonal moments to construct debiased machine learning estimators for intergenerational income mobility measures under plausible missing-at-random and testable independence assumptions. I apply this framework to estimate the intergenerational elasticity for the U.S. using the Panel Study of Income Dynamics across birth cohorts from 1954 to 1977 with rolling 10-year windows. While existing approaches estimate values between 0.41 and 0.54, the proposed method yields substantially higher estimates ranging from 0.6 to 0.7, averaging 0.64. These results align closely with recent evidence using long time averages over mid-career periods, reinforcing high U.S. intergenerational persistence.
Read the full abstract

Number Theory

The Right Edge of the Zero Set of the Fibonacci Zeta Function

Let $F_1=F_2=1$, $F_{n+2}=F_{n+1}+F_n$, and define the Fibonacci zeta function by $$ Z_F(s)=\sum_{n\ge1}F_n^{-s},\qquad \operatorname{Re}s>0. $$ We determine the exact right edge of the closure of the real parts of its zeros in the half-plane of absolute convergence. If $\sigma_F$ is the unique solution of $$ Z_F(\sigma_F)=4+2\,144^{-\sigma_F}, $$ then $$ \sigma_F=0.743163398726901648\ldots, $$ $Z_F(s)\neq0$ for $\operatorname{Re}s\ge\sigma_F$, while $$ \overline{\{\operatorname{Re}\rho:Z_F(\rho)=0,\ \operatorname{Re}\rho>0\}}=[0,\sigma_F]. $$ The edge is sharp in an almost-periodic sense: zeros occur with relatively dense ordinates near every admissible vertical line. We prove growing-dimensional phase locking near the edge and a Diophantine zero-free cusp, and describe the associated Jessen function and smooth mean vertical zero density. For every partial sum with $N\ge12$ we determine the corresponding exact closure edge $\sigma_N$, prove $\sigma_N\nearrow\sigma_F$, and obtain an exponential asymptotic for $\sigma_F-\sigma_N$. We also derive a finite-core theorem for positive integral Lucas zeta functions, with the Pell zeta function as an explicit example. Finally, using the known meromorphic continuation, we construct a natural $q$-Pochhammer completion that is entire of exact order $2$ and type $\log\varphi/4$.
Read the full abstract

Statistical Mechanics

Anomalous diffusion in porous fractal media

We suggest a model of a diffusive process inside a fractal sponge structure, which is a generalization of the diffusion processes on a comb and fractal mesh structure. The sponge model is considered as the direct product of Cantor sets. It is shown that the corresponding one-dimensional diffusion process is governed by a generalized Fokker-Planck equation with a power-law memory kernel and a position-dependent diffusion coefficient. That is, the fractal structure of the medium induces memory effects and heterogeneity in the transport system. The considered model may be of interest to describe anomalous heat transport in porous fractal media.
Read the full abstract

Cosmology and Nongalactic Astrophysics

Expanding the scope of dark siren cosmology: Inferring the population properties of gravitational wave-hosting galaxies

The field of gravitational wave "dark siren" cosmology is facing a number of challenges surrounding the way in which galaxy catalogue data is incorporated into the analysis. The three biggest challenges are adequately compensating for galaxy catalogue incompleteness, robust treatment of galaxy redshift and source properties, and the as-yet unknown host galaxy weighting model. We present an updated methodology for inferring cosmological parameters using dark sirens and incomplete galaxy catalogues. By switching to a sampled redshift prior, rather than one which is pre-computed on a grid, we can jointly infer parameters which describe the population of gravitational-wave hosting galaxies, alongside the gravitational wave population and cosmological model. Using the cosmological inference software gwcosmo, we reproduce a subset of GWTC-5.0 cosmology results, then extend the analysis to jointly infer the host galaxy weighting model - the first measurement of its kind to do so - leading to an updated measurement of the Hubble constant with $H_0 = {71.9}_{-7.5}^{+9.1}$ km s$^{-1}$ Mpc$^{-1}$ (median and 68% symmetric credible interval). This innovative new method opens the door to solving several of the biggest challenges currently facing the field of dark siren cosmology.
Read the full abstract

Cosmology and Nongalactic Astrophysics

Late-Time Emergence of Dark Energy and Its Interaction with Dark Matter

We present an interacting scenario between dark energy (DE) and dark matter (DM), where DE has an emergent nature, that means, DE was absent in the early universe but it becomes effective only at late times. We consider two specific emergent DE models, one with no free parameters and the other featuring two parameters describing the speed and epoch of the transition. We constrain both scenarios using the cosmic microwave background (CMB) measurements from the Planck 2018 release, baryon acoustic oscillations from DESI DR2, and three different compilations of Type Ia supernovae (PantheonPlus, DES-Dovekie, and Union3). Our analysis indicates that current cosmological probes are not able to tightly constrain the speed of the transition. For both scenarios, the posterior distribution of the interaction parameter is shifted away from zero at more than 95% CL whenever the CMB data are combined with any of these additional probes, with the preferred direction corresponding to a transfer of energy from DE to DM. While CMB alone yields a high value of $H_0$, in agreement with local determinations, this effect is reduced when DESI is added and disappears once supernova data are included. In contrast, the clustering parameter $S_8$ is consistently shifted toward lower values in the combined datasets, and it is correlated with the preference for a negative interaction. However, according to the $\Delta \chi^2_{\rm min}$ and Bayesian evidence, none of the interacting models is favored over $\Lambda$CDM or $w_0w_a$CDM, indicating that the interaction does not rescue these emergent DE models. Our results therefore highlight the limitations of these scenarios in addressing current cosmological tensions, while pointing to the crucial role of future data in further assessing their viability.
Read the full abstract

All authors Sibo Zhang, Weiqiang Yang, Supriya Pan, Olga Mena, Eleonora Di Valentino, Subenoy Chakraborty

Optics

Sub-2-Cycle, Terawatt Pulses via Double-Stage Multi-Pass Cell Compression of an Yb Laser

We report a terawatt-class, sub-2-cycle post-compressed Yb-based laser system operating at kHz repetition rate. 18 mJ, 400 fs pulses delivered at 1 kHz by a commercial Yb:YAG laser are spectrally broadened and temporally compressed in a double-stage multi-pass cell setup featuring an argon-filled Herriott-type cell followed by a helium-filled array-type cell, achieving an overall compression factor of 65. The resulting compressed pulses are measured to have 6.2 fs duration (1.8 optical cycles at 1030 nm) and 12.5 mJ energy, corresponding to 1.4 TW effective peak power. To the best of our knowledge, this constitutes the highest peak power reported to date for a few-cycle, kHz-repetition-rate, post-compressed Yb laser system. Focused intensity measurements yield a normalized vector potential of a0 = 3.0, confirming the applicability of this light source for laser wakefield acceleration.
Read the full abstract

All authors Victor Koltalo, Jaismeen Kaur, Louis Daniault, Antoine Cavagna, Mickael Allemand, Florent Pallas, Antoine Courjaud, Emilien Gontier, Cédric Sire, François Sylla, Rodrigo Lopez-Martens

Statistical Methodology

A Strictly Proper Scoring-Rule Theory for Calibrating Stochastic Car-Following Models

Problem definition: Fixed parameters and inputs in a stochastic simulator induce a distribution over complete trajectories, not one trajectory. Calibration must assess this distribution, including variability and temporal dependence, against observations. Yet stochastic car-following models are commonly calibrated with trajectory-error objectives inherited from deterministic modelling. Methodology/results: We establish a scoring-rule theory of stochastic calibration. Strict propriety requires the data-generating distribution to uniquely minimise expected score. MRMean-I, the average run-wise error, drives separable stochastic spread to zero; MRMean-II, the error of the ensemble-mean trajectory, cannot identify a parameter that changes only spread; and MRMin, the error of the closest simulated run, has a population target that changes with ensemble size. These results are confirmed for stochastic Intelligent Driver Model extensions with additive acceleration noise and random desired headway. We recommend exact maximum likelihood when the correct transition density is available; otherwise, an unbiased simulation-based estimator of a strictly proper score. The energy score meets this requirement and gives the best held-out distributional prediction among the evaluated simulation-based objectives, although both models retain too-narrow bands and miss persistent disturbances. Implications:Strict propriety separates a valid calibration target from parameter identifiability and model adequacy. The theory applies to vector-valued outputs from stochastic transportation simulators; the car-following experiments illustrate its scope.
Read the full abstract

All authors Shirui Zhou, Shiteng Zheng, Junzhe Ding, Rui Jiang, Junfang Tian

Plasma Physics

Impact of ion-beam stopping power on proton-boron fusion yield in the pitcher-catcher scheme driven by ultra-intense laser

The effect of stopping power on proton-boron fusion is investigated for a proton beam propagating through boron plasma. Due to the stopping power, the electron temperature of the boron target rises as the proton beam deposits energy. Consequently, a feedback mechanism becomes significant when the trailing part of the beam propagates into the preheated plasma. By coupling this phenomenon with the proton-boron fusion process, fusion yields are systematically investigated by varying the central energy of the proton beam, as well as the thickness and density of the boron target. It is found that, under the influence of stopping power, the optimal central energy for fusion deviates from the intrinsic 672 keV resonance and shifts to approximately 900 keV. Moreover, the present results are substantiated by particle-in-cell simulations, which provides a valuable reference for subsequent high-yield hydrogen-boron fusion.
Read the full abstract

All authors J. Y. Hua, X. F. Li, J. X. Wang, Y. X. Leng, Y. Tian, R. X. Li

Number Theory

Distribution properties of generalized polynomials

A generalized polynomial is a function defined by an iteration of the operations addition, multiplication and the floor function. Equidistribution results of sequences given by generalized polynomials have been established by Håland and later by Bergelson and Leibman from an ergodic theoretic point of view. In the present paper we show an asymptotic distribution result which can be applied to certain generalized polynomials. In the second part we prove equidistribution results for generalized polynomials along prime numbers, including bounds for the discrepancy.
Read the full abstract

History and Philosophy of Physics

From the Light Quantum to the Photon: The Evolution of a Physical Concept

This work examines the physical and conceptual evolution of the light quantum from Planck's blackbody theory to the theoretical and experimental developments that led to the quantization of the electromagnetic field. The present study shows that the decisive transition occurred in Einstein's quantum theory of radiation (1916-1917). Absorption, stimulated emission, and spontaneous emission were formulated as elementary probabilistic mechanisms whose statistical balance alone reproduces the blackbody spectrum. In particular, spontaneous emission requires the emission of a light quantum, thereby implicitly proving its physical necessity before its theoretical status was clarified. At the same time, the already existing term photon began to acquire a stable usage following Lewis's 1926 proposal and became increasingly associated with Einstein's light quantum. By the mid-1920s, the central problem had shifted from whether light quanta were physically required to how radiation could be incorporated into the emerging quantum-mechanical formalism. This transition marks a key stage, illustrating how initial debates about the existence of light quanta gave way to their integration into a comprehensive theoretical structure. The resulting asymmetry between the novel quantum description of matter and the still-classical description of radiation, called into question by the phenomenon of spontaneous emission, identifies the physical problem that led to the quantization of the electromagnetic field.
Read the full abstract

All authors Aaron Collavini, Valentina Bologna, Francesco Longo, Stefano Ansoldi, Fulvio Parmigiani

Computer Vision

Temporal Residual Neural Radiance Fields for Monocular Video Dynamic Human Body Reconstruction

In the field of computer vision and graphics, high-quality reconstruction of the human body in static scenes has been achieved in recent years by a single multilayer perceptron (MLP) in a number of approaches. However, MLPs have capacity limitations, requiring substantial training time and computational resources for dynamic scene reconstruction. And the quality of reconstruction is significantly constrained. This paper proposes a method for effectively processing complex spatiotemporal signals in dynamic scene human 3D modeling. The proposed method uses Temporal Residual Neural Radiance Fields to achieve novel view rendering and new pose synthesis of human bodies.To address the problem of representing temporal signals in video sequences, we construct a temporal residual field which is not related to the MLP architecture. Secondly, to improve reconstruction efficiency, we propose an integrated approach that reduces trainable parameters and accelerates rendering, thereby enhancing the network's feature representation capability. Finally, we design a multi-dimensional loss function to accurately measure the loss between predicted and actual spatial pixel values. The experimental results show that our proposed approach improves the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) accuracy metrics compared to the latest representative methods. It maintains similar accuracy to Anim-NeRF and Neural Body while achieving a nearly 780-fold increase in time efficiency.
Read the full abstract

All authors Tianle Du, Jie Wang, Xiaolong Xie, Wei Li, Pengxiang Su, Jie Liu

Human-Computer Interaction

MIVAIS: A Study Environment for Multi-Agent Mixed-Initiative Visual Analytics Applications

Mixed-initiative Visual Analytics (VA) systems empower human users by interleaving human intuition with software agents and their machine intelligence. However, the development and rigorous evaluation of such systems remain constrained by engineering overhead. Developers must, e.g., implement complex, low-level state synchronization to manage asynchronous agent behaviors, while researchers struggle to capture the multimodal provenance required to study and evaluate human-AI collaboration. We present MIVAIS, a dual-layered research platform designed to abstract the structural complexities of mixed-initiative VA. First, it contributes a computational Infrastructure that standardizes human-software agent interaction, state synchronization, and communication between the agents. Second, it provides a declarative Study Environment that automatically logs multimodal human-AI telemetry - including application/system state, screen capture, audio, and additional sensor data - enabling seamless, in-situ user studies and post-session analysis. We technically validate our infrastructure by replicating three state-of-the-art systems (Podium, Voyager 2, and ProactiveVA). Furthermore, we evaluate the framework's expressiveness and efficiency through expert case studies with HCI and VA researchers, demonstrating how MIVAIS effectively lowers the barrier to prototyping and evaluating intelligent, co-adaptive interfaces.
Read the full abstract

All authors Tobias Stähle, Simon Schneider, Rita Sevastjanova, Mennatallah El-Assady

Category Theory

From Matrices to Morphisms II: Examples of Computational Categories in MATLAB and Octave

This article develops the notion of a computational category, motivated by the observation that many mathematical structures occurring in scientific computing already admit concrete algorithmic representations. Finite sets, finite-dimensional vector spaces, relations, and graphs may all be represented through elementary MATLAB and Octave data structures such as indexing vectors, matrices, logical arrays, and sparse matrices. The central idea is that a category may often be described through a category of canonical representatives indexed by the natural numbers. This viewpoint leads naturally to representative categories whose objects are dimensions and whose morphisms are finite computational data structures. The resulting framework unifies several examples, including matrix categories, index categories, categories of column spaces, row-based models of finite sets, and categories combining finite subsets with finite-dimensional vector spaces through mixed morphisms. The paper also introduces computational categories with distinguished non-finite representatives, extending the finite setting while retaining concrete coordinate descriptions. Throughout, categorical constructions are interpreted algorithmically and implemented using standard MATLAB and Octave operations. The work contributes to a computational perspective on category theory and, conversely, to a categorical understanding of matrix-oriented scientific computing.
Read the full abstract

Artificial Intelligence

A Tree-based RAG Framework for Evidence-Intensive QA via Adaptive Planning and Topology-Aware Evidence Gathering

Recent structured RAG methods leverage tree- or graph-based reasoning structures to improve multi-hop QA. However, they face key limitations in evidence-intensive QA, where answering a question requires synthesizing information scattered across dozens or even hundreds of documents: structural rigidity, which limits adaptive reasoning expansion, and topology-ignorant evidence gathering, which prevents effective integration of evidence across different reasoning nodes. To address these issues, we propose APT-RAG, an Adaptive Planning and Topology-aware evidence gathering RAG framework. Adaptive planning dynamically expands the reasoning structure based on question dependencies and evidence requirements, while topology-aware evidence gathering improves evidence coverage through sibling evidence reuse, direct retrieval, and evidence aggregation from child nodes. We further introduce evidence-guided batched answer generation to reduce significant generation overhead in evidence-intensive QA. In the experiments on evidence-intensive QA benchmarks, APT-RAG outperforms existing structured RAG methods. Our code is available at https://github.com/hyudsl/APT-RAG.
Read the full abstract

All authors Songeun Lee, Kyungjin Min, Injae Na, Suyeong Lee, Chiyoung Kim, Woohwan Jung

Number Theory

On the Divisibility Relation $\sigma(n)\mid\sigma(n+h)$ and a Generalized Erdős–Sierpiński Conjecture

For each fixed positive integer $h$, we study the divisibility relation $\sigma(n)\mid\sigma(n+h)$. We isolate an explicit regular family arising from integral quotients of shifted abundancy indices and show that the complementary set satisfies a subexponential saving; in particular, the number of solutions up to $x$ is $O_h(x/(\log x)^2)$. We also study the proportionality equation $\sigma(n+h)=\lambda\sigma(n)$. For every fixed nonzero integer $h$, uniformly for all real $\lambda>0$, the number of solutions up to $x$ is $O(x/\sqrt{\log\log\log x})$, with an absolute implied constant once $x$ exceeds an $h$-dependent threshold. Finally, we give an explicit family which, under Schinzel's Hypothesis $H$, produces infinitely many solutions of $\sigma(n+1)=2\sigma(n)$; the Bateman–Horn conjecture yields a precise asymptotic for the number of members of this family up to $x$. We conjecture that $\sigma(n+h)=k\sigma(n)$ has infinitely many positive integer solutions for every fixed $h,k\ge1$.
Read the full abstract

Combinatorics

A Short Combinatorial Proof of the Pons-Batle Identity for Counting Tree-Child Networks

Tree-child networks are a useful class of binary phylogenetic networks. The Pons–Batle identity (Pons and Batle, Scientific Reports, 2021) states that the number $a_{n,k}$ of tree-child networks with $k$ reticulations on $n$ taxa satisfies \[ a_{n,k}=(n-k+1)a_{n,k-1} +\frac{n(2n+k-3)}{n-k}a_{n-1,k}. \] In this paper, we present a short combinatorial proof of this identity.
Read the full abstract

Artificial Intelligence

Global to Local: Topology-Preserving Adaptive Graph Pooling via Granular-Ball

Graph pooling aims to compress the graph, including both node embeddings and their underlying topological patterns, into a more compact representation. Previous works focus primarily on the overly fine-grained representation of nodes, progressively coarsening the graph by removing nodes or merging them into clusters, thus neglecting the global-to-local patterns and adaptive granularity of the graph's topological structure. In the real scenario, graphs as a whole can be considered the coarsest level of granularity, encapsulating the global topological structure, with progressively finer-grained local topological structures represented from top to bottom. This process continues until the adaptive granularity for each subdomain is reached. To this end, we propose a novel Topology-Preserving Adaptive Graph Pooling (TPAGP) method that dynamically partitions graphs into granular balls by integrating node features and topological information, enabling the generation of multi-granularity representations that effectively capture both local and global structural patterns. Additionally, we design a multi-granularity graph network model that facilitates feature interaction and optimization across different granularities, significantly enhancing performance in graph classification tasks. Experimental results demonstrate that TPAGP outperforms existing pooling methods across various benchmark datasets, effectively mitigating information loss caused by fixed-granularity strategies.
Read the full abstract

All authors Sen Zhao, Gaojie Xu, Shuyin Xia, Yifan Guan, Yi Liu, Yi Wang, Wei Wang

Chemical Physics

Operando imaging of intercalation memory in MXenes

Ion intercalation enables reversible control of charge, structure and function in layered solids. Yet layered materials contain nanosheets with different thicknesses and stacking, whose intercalation pathways are averaged out in ensemble measurements. This becomes critical under Angstrom-scale confinement, where ion transport couples to solvent reorganization and host deformation, leaving the origins of kinetics, reversibility and activation unresolved. Here we combine operando interferometric scattering and interference reflection microscopy to resolve proton-driven dynamics in individual single- and multilayer Ti3C2Tx MXene flakes and show that even a few stacked layers introduce a significant kinetic barrier. Cycling then separates nanosheets into three persistent behaviors: reversible monolayers, restacked multilayers with incomplete recovery and delayed deintercalation, and coherently stacked multilayers that self-stabilize through reversible folding. Operando synchrotron X-ray diffraction shows this memory at electrode scale, where the interlayer structure converges towards a reproducible state. Our findings reveal cycling-induced structural memory that governs subsequent intercalation in layered MXenes.
Read the full abstract

All authors Franz Groebmeyer, Mohsen Beladi, Christoph G. Gruber, Ruocun Wang, Pol Salles, Nhu Quynh Nguyen, Jakub Drnec, Yury Gogotsi, Emiliano Cortes

Computer Vision

RefDiT: Local Attribute Guidance in Reference-Based Image Generation

Personalization models generate new images guided by a few subject references, while style transfer methods aim to produce images aligned with a global style derived from a reference image. Recent approaches perform well when the reference image contains a single object, effectively capturing a global style that encompasses all implicit attributes. However, when applied to complex real-world scenes containing multiple objects with distinct attribute characteristics, these methods, due to their global-level guidance, fail to localize relevant elements in the reference image. The global guidance restricts their ability to generate new images based on the local attributes in the reference image. Moreover, existing methods typically employ a single identifier token to capture all details from the reference, resulting in a lack of individual, attribute-level control. Motivated by these limitations, we propose RefDiT, a novel framework for reference-guided image generation. RefDiT takes as input a reference image, a text prompt, and an optional user-provided guidance context. RefDiT employs local region guidance using the attributes of local elements. It constructs an attribute-aware conditioning signal from the reference image by performing attribute-level decomposition of the identifier token and performs context adjustment in the inference prompt to train low-rank adapter (LoRA) blocks of a diffusion transformer (DiT)-based generative model. RefDiT learns the correspondence between identifier tokens and local regions in the reference image, enabling more effective local guidance.
Read the full abstract

Sound

One-Stage Multi-Task Instruction-Guided 3D Spatial Audio Editing

Spatial audio editing modifies an existing soundfield according to a user's instruction while preserving the rest of the scene. Unlike conventional audio editing, it must reason jointly about audio events, spatial information, dynamic changes, and environmental information in first-order Ambisonic (FOA) waveforms. Existing language-guided editors mainly target conventional audio or rely on sequential operations, and therefore do not directly support one-stage editing for complex 3D spatial instructions. We present SwanWeave, the first one-stage multi-task framework for instruction-guided 3D FOA spatial audio editing. We build paired FOA supervision from open-source speech and sound-effect corpora using controllable room simulation, covering more than ten single-operation and compound tasks across the four editing axes. To handle this heterogeneous edit space, SwanWeave uses Spatial Edit Mixture-of-Experts (SE-MoE) with dual-level routing, selecting task-aware expert combinations for compound instructions and frame-level routed/null experts for local edit decisions. We further introduce Spatial Preference Optimization (SPO), a Direct Preference Optimization (DPO)-based alignment objective with edit-specific negative targets, and adopt staged training to improve natural-language grounding. Experiments show that SwanWeave achieves better editing quality than existing general audio editors and spatial audio baselines across all tasks. Spatial audio editing demos can be found at https://swanaigc.github.io/#swanweave, code can be found at: https://github.com/MM-Speech/SwanWeave.
Read the full abstract

All authors Ke Lei, Chenyuhao Wen, Yu Zhang, Wenxiang Guo, Changhao Pan, Sashuai Zhou, Yongshi Li, Ruiqi Li, Ruofan Hu, Haorui Xu, Xiang Yin, Zhou Zhao

Functional Analysis

Dimension-free $H^\infty$-calculus of angle $<\pi/2$ for UMD-valued Ornstein–Uhlenbeck operators

Let $1<p<\infty$, let $X$ be a UMD Banach space, let $1<p<\infty$, and let $L_d$ be the generator of the Ornstein–Uhlenbeck semigroup $(P_d(t))_{t\geq 0}$ on $L^p(\mathbb R^d,\gamma_d;X)$. We prove that the operators $-L_d$ are $R$-sectorial with a common angle strictly smaller than $\pi/2$ and with bounds independent of $d$. Combining this with the Hieber–Prüss transference theorem and the Kalton–Weis angle comparison, we deduce that the operators $-L_d$ admit bounded $H^\infty$-calculi of a common angle strictly smaller than $\pi/2$, again with dimension-free bounds. The corresponding Walsh $R$-analyticity estimates are proved first and transferred to the Ornstein–Uhlenbeck setting by a central limit argument.
Read the full abstract