Peer Revenue Publish or profit

Papers

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

All of arXiv

Thursday 3 September

Computer Vision

ArtSociety: Multi-Agent Multimodal Collaboration for Art Emotion Understanding

The AffectiveArt Multidimensional Art Emotion Understanding task asks to jointly predict an artwork's fine-grained emotion (12 classes, 1549:1 head-to-tail ratio), binary valence/arousal, and five attribute-grounded descriptions – sub-tasks that exhibit strong empirical trade-offs, so the single-model solutions we tried do not jointly optimize all of them well. We present ArtSociety, a multi-agent framework that assembles heterogeneous multimodal experts – a DINOv2-Giant vision agent (A1), a scene-grounded CoT fine-tuned MLLM (A2), and three closed-source reasoning agents (A3-A5) – and coordinates them with two training-free controllers: (i) a rare-class-aware voting arbiter that lowers the agreement threshold for tail emotions, exploiting decorrelated error patterns across agent families; and (ii) a description-first reasoning agent whose DESCRIBE-then-CLASSIFY chain of thought forces visual evidence before label commitment, yielding near-perfect grounded descriptions. A task-routing policy directs the hard emotion task to the full five-agent ensemble while assigning the near-saturated valence/arousal and generative description tasks to the single strongest reasoning agent. On the official test set (1,000 artworks), ArtSociety achieves an Overall Score of 0.8870 (Classification 0.7789, Description 0.9952). An eleven-variant ablation study reveals that, once method and scale saturate at around 0.76, the decisive gains come from agent collaboration and data-side supervision – a 30B MoE model trained on older data does not outperform an 8B model trained on better data. Code is available at https://github.com/swordlidev/ArtSociety
Read the full abstract

All authors Jian Li, Fanfan Ji, Jinxiang Lai, Ying Tai, Jian Yang, Xiao-Tong Yuan, Chengjie Wang, Yabiao Wang

Computer Vision

Abstract-LoRA: Unlocking Single-Image Style Transfer through Targeted U-Net Block Training

Diffusion models represent one of the most advanced paradigms in generative modeling. Leveraging their development, a growing number of style transfer methods based on diffusion models have been proposed. However, among these methods, multi-image style transfer approaches that require at least five to ten style examples tend to achieve more satisfactory results. Single-image methods, by contrast, often struggle with either insufficient content preservation or inadequate style fidelity. This greatly limits style extraction from scarce artworks and undermines their artistic value. To address this, we propose Abstract-LoRA, a method that pushes the boundaries of single-image style transfer through lightweight LoRA training on specific U-Net blocks in diffusion models. Specifically, our work is inspired by B-LoRA, a style transfer method that achieves basic style-content disentanglement by training specific U-Net blocks. However, it suffers from a critical limitation: the inability to capture complex backgrounds. Building upon B-LoRA, our method conducts a more refined analysis of U-Net blocks, employing additional U-Net blocks and clustering-based abstraction of style images to better disentangle and balance style and content. Extensive experiments demonstrate that our proposed method not only generates visually more harmonious and satisfying artistic images but also quantitatively improves the preservation of both style and content in the final outputs.
Read the full abstract

General Physics

Cosmological Constraints on Interacting Scalar-Field Dark Matter and Early Dark Energy mediated by a Kinetic Kernel

In this work, we build a general framework for constraining interacting scalar fields with cosmological observations, applicable to a broad family of scalar–field potentials and interaction kernels. As a representative example, we derive constraints on a model in which dark matter is described by a scalar field $\phi$ with a quadratic potential and dark energy by a scalar field $\psi$ with the Albrecht–Skordis potential, considering both the uncoupled case and a more general scenario in which the exchange of energy occurs through an interaction kernel proportional to the kinetic energies of the scalar fields $Q \propto \dot{\phi}^2 \dot{\psi}^2$. The scalar field parameters are constrained by using Planck 2018 CMB data, DESI BAO measurements, Ly-$\alpha$ forest constraints, and SNeIa observations, and we find that the scalar field parameters are driven toward their $\Lambda$CDM-like limits; the early dark energy fraction is largely suppressed, and the SFDM mass is bounded to values for which the field behaves close to cold dark matter. In the interacting case, the results show no significant preference for the transfer of energy between dark energy and dark matter at 95% C.L. In general, the model keeps $H_0$ close to the Planck-$\Lambda$CDM range and provides best-fit $\chi^2$ values comparable to those of the corresponding $\Lambda$CDM fits.
Read the full abstract

General Physics

Observational constraints on Diffusion Cosmologies in Unimodular Gravity from DESI DR2

Diffusion functions in unimodular gravity induce a dynamical effective cosmological constant, providing an appealing framework in light of recent DESI observations. In this work, we investigate the observational viability of a general class of diffusion models using baryon acoustic oscillation measurements from DESI DR2, Type Ia supernova compilations (Pantheon+, DES-Dovekie, and DESY5), and Cosmic Chronometer data. We find that diffusion models systematically achieve lower best-fit $\chi^{2}$ values than $\Lambda$CDM across all dataset combinations considered, indicating a modest but persistent improvement in goodness of fit. Nevertheless, the reduction in $\chi^{2}$ is insufficient to offset the larger parameter space, leading standard information criteria to favor the simpler $\Lambda$CDM model. Despite this result, all dataset combinations consistently prefer a nonvanishing diffusion contribution corresponding to approximately $20\%$ of the present dark-energy budget, with a posterior probability $P(f_Q>0.05) =96.78\%$. From an observational perspective, diffusion models therefore remain a viable extension of the standard cosmological scenario and motivate the exploration of simpler diffusion parameterizations, particularly in light of upcoming high-precision cosmological surveys.
Read the full abstract

General Physics

Cosmological constraints on the interacting viscous dark matter and decaying vacuum energy models

We investigate an interacting dark sector scenario in which a deviation from the standard dilution law of cold dark matter is parametrized by a perturbative correction $\epsilon$, thereby inducing a dynamical vacuum component. We extend this framework by allowing dark matter to behave as an imperfect fluid with bulk viscosity. Within the Eckart formalism, we consider a power-law dependence of the bulk viscous coefficient on the dark matter energy density which yields an effective density scaling for the viscous pressure. We obtain the parameter constraints using multi-nested sampling with Dynesty, employing DES 5YR Type Ia Supernovae, DESI Baryon Acoustic Oscillations, Cosmic Chronometers, the local $H_0$ measurement from H0DN, and Cosmic Microwave Background distance priors. Redshift Space Distortion measurements from SDSS-IV and the DES weak-lensing constraint on $S_8$ are included to probe structure formation. We also examine the thermodynamic consistency of the cosmological scenarios and find that they satisfy the Generalized Second Law of thermodynamics. We find that bulk viscous effects can modify the late-time interaction dynamics, while their contribution is suppressed once early-Universe information is included. These results provide observational and thermodynamic constraints on departures from the standard cold dark matter evolution and a constant vacuum energy.
Read the full abstract

General Relativity and Quantum Cosmology

Constraints on Scalar–Tensor–Vector Gravity Theory Parameters Inferred from Quasiperiodic Oscillations

We study circular geodesics of neutral test particles in the static charged solution of Scalar–Tensor–Vector Gravity (STVG), and we use the twin kilohertz quasiperiodic oscillations (QPOs) of twelve accreting compact objects to bound its parameters. Starting from the effective potential we obtain closed forms for the specific energy, the specific angular momentum and the Keplerian angular velocity, and from these the radial and vertical epicyclic frequencies. The horizon, photon sphere, shadow radius and innermost stable circular orbit are given in closed form as well, and each one reduces to the RN and Schwarzschild value in the appropriate limit. Null geodesics are integrated numerically and show how the enhanced coupling widens the capture cross section while leaving the logarithmic divergence of the deflection angle at the photon sphere intact. Within the RP model we then run Metropolis–Hastings MCMC simulations on the QPO pairs of eight neutron stars and four microquasars, comparing the Schwarzschild spacetime, the RN spacetime, STVG with vanishing charge and the full STVG solution. The best fits are obtained mainly for the charged solutions in the neutron star sample, while all four models describe the black hole sample equally well. We show that the orbital dynamics depends on the mass $M$, the coupling $\alpha$ and the charge $Q$ only through the two combinations $\Meff=(1+\alpha)M$ and $\Qeff^{2}=(1+\alpha)(\alpha M^{2}+Q^{2})$, which accounts for the multimodal posteriors found in the neutron star sample and sets a model-independent limit on what QPO timing alone can measure. The astrophysical implications of these bounds, in particular for inferred masses above the Tolman–Oppenheimer–Volkoff limit, are discussed in detail.
Read the full abstract

All authors Marco Muccino, Kuantay Boshkayev, Binay Prakash Akhouri, Izzet Sakall{\i}, Yassine Sekhmani

General Relativity and Quantum Cosmology

Entropy Obstruction to Closed Semiclassical Bounces

We prove a finite-$G\hslash$ singularity theorem for semiclassical spacetimes with compact Cauchy slices. Let a compact Cauchy slice be divided by a compact surface into regions $B$ and $C$. Suppose that $B$ is conditionally hyperentropic, $H_{\max,\mathrm{gen}}^\varepsilon(BC|C)>0$, that the future-inward null boundary from the dividing surface toward $B$ is a discrete max lightsheet, and that $C$ is robustly quantum trapped. Assuming discrete max-focusing and a regular semiclassical endpoint for a closing lightsheet, the future Cauchy development of $B$ contains an incomplete null generator. This parallels entropy singularity theorems in which hyperentropy is combined with the existence of an inward lightsheet, while robust quantum trapping supplies the additional finite-$G\hslash$ local obstruction needed here. In a closed Friedmann universe, the theorem excludes a controlled semiclassical bounce when a contracting hemisphere lies on such a lightsheet and contains more independent information than its boundary can support.
Read the full abstract

Neural and Evolutionary Computing

CompEvo: Competition-Induced Evolution for Multi-Agent in News-Driven Time Series Forecasting

News-driven time series forecasting uses evolving textual events together with historical observations to predict future values, supporting applications such as market risk monitoring and resource scheduling. In multi-agent settings, two challenges still remain. The first is degeneration of thought, where agents converge to similar evidence-seeking behaviors. The second is insufficient theoretical grounding, where strategy updates are often heuristic and lack a principled formulation. To address the above challenges, we propose CompEvo, a competition-induced evolution framework for multi-agent news-driven time series forecasting. For theoretical grounding, we introduce an evolutionary game formulation to guarantee equilibrium existence and optimization convergence. Building on this formulation, we construct a trainable multi-agent evolution framework that integrates strategy execution, fitness-based differentiable selection, and competition-induced strategy evolution. CompEvo enables heterogeneous agents to explore diverse news evidence, converts forecasting feedback into differentiable influence weights, and evolves agent strategies under competitive pressure to preserve effective logic while maintaining diversity. Experiments on four real-world datasets show that CompEvo reduces RMSE by 27.3% and MAPE by 26.2% on average over strong baselines. Further analysis indicates that CompEvo successfully maintains diverse and specialized agent behaviors.
Read the full abstract

All authors Yuxuan Zhang, Yangyang Feng, Yong Guan, Daifeng Li, Kexin Zhang, Junlan Chen, Bowen Deng, Jun Liu, Zehua Zeng

Numerical Analysis

Geometry Parameterisation via a Structural Modal Basis for Aerodynamic Shape Optimisation

Aircraft outer mould lines are often heavily constrained early by payload requirements, manufacturability limits, and low-observability considerations. The remaining aerodynamic design space is typically small, highly constrained, and difficult to represent using a concise and physically meaningful set of design variables. A structural modal parameterisation method (MPM) is introduced for such problems. The method constructs a tunable pseudo-structure and solves its eigenvalue problem to obtain a modal basis, from which a selected subset of modes parameterises the geometry. Boundary conditions, stiffness distribution, density, thickness, and added masses are treated as intentional basis-design variables that shape the admissible deformation space prior to optimisation. The resulting design coordinates are independent of any specific flow solver, making the approach broadly applicable. For tightly constrained aerodynamic shape optimisation problems, it provides a practical balance of parameter conciseness, robust constraint awareness, and geometric flexibility.
Read the full abstract

All authors Alwin Wang, Robert Carrese, Luca Brown, Joel Collins, Pier Marzocca

Statistical Methodology

Random Jump Intensities and Bernstein Density Estimation in Ergodic Basic Affine Jump-Diffusion Processes

This paper develops a random-effects extension of the ergodic basic affine jump-diffusion (BAJD) model for a population of independent trajectories with individual unobserved jump intensities and common structural parameters. Under continuous-time observation, each intensity is estimated by the empirical jump frequency. We establish the strong consistency and stable mixed-normal limit of this estimator as the observation horizon tends to infinity. The estimated intensities are then used as pseudo-observations to construct a Bernstein-polynomial estimator of their common density. Its bias, variance, mean integrated squared error, and pointwise asymptotic normality are derived under a sequential asymptotic framework. The theory is developed for general positive jump-size distributions with finite activity and specialized to the Gamma$(k,\lambda)$ family, including the exponential and Erlang cases. For this family, we derive a pooled estimator of the common rate parameter and establish its consistency, asymptotic normality, and first-order conditional asymptotic independence from the individual intensity estimators. We also propose a consistent plug-in estimator of the population stationary mean. The methodology is illustrated through simulations and an empirical application to financial realized-volatility data.
Read the full abstract

Populations and Evolution

Spread of Chronic Wasting Disease under Stochastic Environmental Conditions and its Control using Deep Reinforcement Learning

Chronic wasting disease (CWD) is a fatal prion disease affecting deer, elk, moose, reindeer, muntjac, and other cervids. Because free-ranging cervid populations face environmental variability and randomness, deterministic models may miss important dynamics like stochastic fade-out. We develop a stochastic Susceptible-Infectious-Environmental model using differential equations with reflection to ensure the susceptible class remains non-negative. We examine how environmental variability influences cervid populations as CWD pressure and control measures increase. For the deterministic model, we derive the basic reproduction number as the sum of direct and environmental contributions, showing the endemic phase arises at R0=1. For the stochastic system, we establish local well-posedness, positivity, and the disease-free law. The top Lyapunov exponent for invasion remains unaffected by reflection. We evaluate CWD mitigation using a deep reinforcement learning agent trained with Proximal Policy Optimization in a hybrid action space, comparing hunting, decontamination, and combined strategies. In the deterministic case, hunting alone can control the disease but reduces the population by about 58%, while decontamination requires sustained effort. The combined policy more than doubles the cervid population and nearly eliminates infection and contamination. In the stochastic case, the policy contains the disease in about 80% of runs, with 10% experiencing large outbreaks; effectiveness decreases as noise increases. Across all scenarios, the agent consistently emphasizes environmental decontamination, the key control method.
Read the full abstract

All authors Wei Yin, Wesley J. Marrero, Kamal Jnawali, Lale Asik, Michael G. Tyshenko, Tamer Oraby

Distributed Computing

Toward Sustainable Distributed LLM Inference: A Systems Synthesis and Research Agenda for an Energy-, Carbon-, and Cache-Aware llm-d Control Plane

Large language model (LLM) sustainability is increasingly a serving-systems problem, not only a training problem. In production, energy and carbon impact depend on more than model size: workload shape, batching, key-value (KV) cache reuse, prefill/decode placement, model and accelerator choice, power state, geographic carbon intensity, and service-level objectives (SLOs) all matter. Recent systems papers study many of these factors separately. This paper connects those results and asks a practical engineering question: what do they imply when the decision point is a distributed inference control plane such as llm-d? The contribution here is synthesis, not a new set of benchmark results. Reported performance, energy, carbon, and cost improvements remain the results of the cited papers and systems. I group the literature into recurring design patterns and use those patterns to sketch a Sustainable Inference Control Plane (SICP) for llm-d. The proposed control plane would consider latency, energy, carbon, cache reuse, serving cost, and quality when routing and scaling, while keeping TTFT/TPOT SLOs as hard constraints. I also outline an evaluation framework based on SLO-satisfied goodput per joule and per gram CO2e, together with a reproducible experimental plan. The main observation from connecting the literature is that sustainable LLM inference is unlikely to come from one "green" model or one accelerator; it is more naturally treated as a control problem across model, phase, cache, hardware, replica, region, and time.
Read the full abstract

Cryptography and Security

Constructions of complete permutations over $\mathbb{F}_q^n$

Complete permutation polynomials play an important role in cryptography, particularly in the design of cryptographic primitives such as the Lai–Massey scheme and S-boxes. We generalize a result of Sun, Li, Guo, and Qu (2021) by characterizing the complete permutation behavior of the mapping $\Psi(X)=M(X+\psi(AX))$ over $\mathbb{F}_q^n$, where $\mathbb{F}_q$ is a finite field of $q$ elements with $q$ being a prime power, $M\in GL(n, \mathbb{F}_q)$, $GL(n, \mathbb{F}_q)$ is the general linear group of order $n$ over $\mathbb{F}_q$, $A_{m \times n}$ is a full-rank matrix over $\mathbb{F}_q$, and $\psi=(\psi_1,\psi_2,\ldots,\psi_n)$ with each component function $\psi_i:\mathbb{F}_q^m\to\mathbb{F}_q$. Furthermore, we establish criteria for the permutation and complete permutation properties of the mapping $F(X)=T(X+B^tf(AX))$ over $\mathbb{F}_{q}^n$, $f: \mathbb{F}_{q}^{m} \rightarrow \mathbb{F}_{q}^{n-m}$, $T \in GL(n, \mathbb{F}_q)$, $A_{m \times n}$ and $B_{(n-m)\times n}$ are full-rank matrices over $\mathbb{F}_q$, $B^t$ represents the transpose of the matrix $B$, and $0<m<n$ are integers. These results also generalize an earlier result of Gravel and Panario (2023), who showed that any arbitrary function $f$ from $\mathbb{F}_q^m$ to $\mathbb{F}_q^{\,n-m}$ can be extended to a bijection over $\mathbb{F}_{q}^n$ through the mapping $F(X)=T(X+B^tf(AX))$, under the condition $AB^t=0$. Here we do not impose the restriction that $AB^t=0$.
Read the full abstract

All authors Sartaj Ul Hasan, Ramandeep Kaur, Hridesh Kumar, Daniel Panario, Qiang Wang

Software Engineering

Look Before You Prompt, and After: Scaffolding Human-AI Collaboration in Software Tutorial Creation

With LLMs, creating software tutorials now involves steering the model's output and shaping it into a coherent, accurate learning resource, yet existing LLM tools offer writers little support for this work. By analyzing interviews with technical writers ($N=17$), we identify three requirements for how they assemble and structure multiple LLM responses, curate the context the model uses, and verify the generated content. We designed a tool called dBlocks with the following features: blocks to scope content, a context manager to edit context, and inline execution to verify code. Following a human-centered design method, we iteratively refined the design through a user study ($N=5$). In a within-subjects lab study ($N=16$) comparing dBlocks with participants' preferred workflows for LLM-assisted authoring, participants reported significantly higher confidence in the tutorials they produced with dBlocks. In addition, the tool reduced friction in verification, with writers verifying code as they drafted rather than deferring or skipping it, and helped them avoid searching long chat histories by scoping their work into blocks that kept each tutorial section and its LLM conversation together. More broadly, our work offers implications for tools that scaffold human-AI collaboration in SE workflows and shows how human-centered design can guide the development of LLM-integrated tools.
Read the full abstract

General Mathematics

A simple method of computing Riemann-type integrals

The "Riemann-type integrals" of the title are not a single new integral, but the family of classical constructions – Riemann, Riemann-Stieltjes, Young-Kondurar, complex-analytic, and Itô – unified by one elementary method. We give a simple argument computing the Riemann integral of a polynomial directly, establishing the fundamental theorem of calculus for polynomials and then continuous functions. The same idea extends to the real line, the complex plane, the multidimensional case, the Riemann-Stieltjes integral, and the Itô integral of a polynomial of a Wiener process (more generally, a continuous semimartingale), recovering the classical Itô formula. The contribution is methodological: a single algebraic telescoping identity, paired with an estimate on a quadratic remainder term, drives every case. What changes is only the remainder's fate: it vanishes identically for polynomials on $\mathbb{R}$; it vanishes in the limit under a quadratic-variation condition $Q(\pi)\to0$ on $\mathbb{C}$ and for Hölder paths; and for a continuous semimartingale it survives as the quadratic variation producing the Itô correction – except for a planar Brownian motion, or more generally a conformal martingale, where an algebraic cancellation inside the complex square removes it again. We regard this last point – that the Itô correction and its cancellation in the conformal case are two instances of the same quadratic remainder, not unrelated facts – as the paper's most distinctive observation. The manuscript has three parts: Part I develops the core method on $\mathbb{R}$ and $\mathbb{C}$; Part II extends it deterministically to the multidimensional, Riemann-Stieltjes, and Hölder (Young-Kondurar) cases; Part III extends it to the stochastic setting.
Read the full abstract

Machine Learning (Statistics)

Rollcast: Proper-Score Gated Rolling Anchors for Adaptive Probabilistic Time-Series Forecasting

Rollcast is a probabilistic forecasting method for univariate time series that combines a compact set of rolling statistical anchors rather than relying on a single global model. Rolling means, medians, extrema, regression endpoints, and quantiles define candidate forecast locations and a representation of the current state. A state-dependent softmax gate learns anchor probabilities by minimizing negative log predictive density, while residual distributions retrieved from similar historical states provide local uncertainty. Recursive simulation propagates the resulting mixture over multiple forecast horizons. The method is evaluated in a Monte Carlo study covering eight data-generating processes, including autoregressive, random-walk, local-trend, threshold, regime-switching, stochastic-volatility, heavy-tailed, and variance-break dynamics. Across 2,000 independent fitted series, Rollcast is compared with the true conditional predictive distribution generated by an oracle simulator. Overall empirical coverage is 86.2% for nominal 90% intervals and 91.5% for nominal 95% intervals. Predictive intervals are on average 13.6% wider than the oracle at the 90% level, while CRPS is 14.4% higher than oracle CRPS. Performance is closest to the oracle under autoregressive, threshold, stochastic-volatility, heavy-tailed, and variance-break dynamics, while local trends and regime switching are more challenging. The results indicate that Rollcast can construct competitive probabilistic forecasts from simple, interpretable local summaries, while also identifying limitations in calibration and recursive uncertainty propagation.
Read the full abstract

Physics and Society

Response-Aware Coarse-Graining in Summary-Mediated Opinion Dynamics

Interacting populations increasingly respond to compressed representations generated from their own collective state. We ask which information such an endogenous coarse-graining must preserve in order to reproduce the collective dynamics. The question has two complementary answers for continuous opinions with attraction, indifference, and repulsion. Inside any strict $K$-bloc response cell where every represented stance attracts every bloc, the dynamics reduce exactly to translation and disagreement. The represented distribution's shape cancels from all disagreement modes, which form a homogeneously anchored weighted signed-Laplacian system, while represented-mean error acts only on translation. This reduction yields a geometry-dependent stability boundary and shows that prevalence bias produces secular drift at marginal stability only through its projection onto the neutral eigenspace. Outside that common linear-response regime, threshold geometry makes distributional information observable. The induced force discrepancy is a behavior-defined integral probability metric; restricted force probes identify generic finite empirical discussions, and for finitely supported candidates uniform force error controls total-variation and Wasserstein error while support reduction incurs a nonzero distortion floor. A cumulative-discrepancy characterization further explains why transport-small summaries can remain behaviorally inaccurate at response thresholds. Mode-resolved integrations, projected particle simulations, finite-size ensembles, and threshold-crossing experiments test these two regimes. As a deployment consequence, broadcast summary error survives as collective noise whereas independently personalized error self-averages. The model is stylized and not empirically calibrated; generated discussion summaries motivate the general problem of endogenous, response-aware coarse-graining.
Read the full abstract

Computational Engineering

LLM-Aided Design for Manufacturing: A Multi-Agent System for Intent-Preserving Redesign of CAD for Improved Manufacturability

We introduce autonomous, intent-preserving Design for Manufacturing (DFM) redesign of CAD parts: given an engineer's CAD model, the method returns a variant that is easier to manufacture without losing its design intent. Generating such a redesign in a single shot is unreliable, since CAD fidelity degrades as parts grow complex; we instead produce it as a sequence of individually verified design transitions. Our DFM-Redesign pipeline realizes this with two coupled agent subsystems driven by a pretrained multimodal LLM: a DFM Reviewer that inspects the current design and proposes one intent-preserving manufacturability improvement at a time, and a CAD Modifier that executes each proposal as an edit to the part's CadQuery program. The CAD Modifier closes a verification loop, compiling every candidate edit and visually checking it against the intended change from multi-view renderings, then re-generating or re-instructing until the edit is accepted or abandoned. Iterating review and verified modification compounds edits into parts more complex than one-shot generators reliably produce, preserves the original intent at each step, and requires no fine-tuning. On a 46-part benchmark scored by chamfer distance to reference geometries, the CAD Modifier reproduces target parts more accurately on average than chain-of-thought single agents given the same tools, and ablations isolate the contributions of the visual review loop and of captioning the design state before each edit. A centrifugal pump casing built from 32 chained transitions illustrates the complexity reachable by compounding verified edits. This is a preliminary report: evaluation of the full review-and-redesign loop, including manufacturability gain and an operational measure of intent preservation, is ongoing.
Read the full abstract

All authors Kojo Welbeck, Xiangyu Shi, Zahra Sadeghi, Qi Zhu, Ping Guo

General Relativity and Quantum Cosmology

The Einstein Tensor from UV-regulated Field Correlations

It is known that the spacetime metric can be recovered from quantum field correlators. We extend this programme by showing that the Einstein tensor can be directly written in terms of UV-regulated field correlations. The same correlation structure also directly determines the Ricci tensor, while higher coincidence derivatives recover the connection and Riemann tensor. We further show that the renormalized stress-energy tensor can be formulated within this framework, up to the standard finite renormalization freedom. These results provide a correlation-based organization of the semiclassical Einstein equations and advance the programme of formulating spacetime geometry in terms of quantum field correlations.
Read the full abstract

Geophysics

Estimating subsurface water mass changes with ground-based gravimetry

Gravitational force is proportional to the mass of an attracting body; therefore, changes in subsurface mass can be detected using gravimetry. In the near subsurface, mass variations are primarily driven by changes in water storage, meaning that gravity measurements provide direct information on variations in water mass. Traditionally, groundwater and soil water dynamics are monitored using point-scale instruments that require direct installation into the subsurface. In contrast, gravimetry is a non-invasive method that integrates signals over a larger spatial extent. Recent advancements in gravimeter technology have significantly improved measurement accuracy and long-term stability under field conditions, thereby expanding their potential for hydrological applications. This review summarizes recent developments in gravimetric instrumentation and provides essential background on gravimeters used for ground-based measurements and their application in hydrology, aiming to facilitate their broader use. In addition, it presents an overview of studies in hydrology and hydrogeology that have successfully applied gravimetry to quantify subsurface water storage changes, estimate hydrological parameters, and constrain numerical models. A perspective on future advancement in gravimetry and resulting potential applications is given.
Read the full abstract

All authors Isabelle Schmidt, Jackson Ang'ong'a, Peter Bauer-Gottwein, Stephan Costabel, Christian Freier, Bastian Leykauf, Tina Lundgren, Miguel Martinez-Dorantes, Achim Peters, Ernst Rasel, Enrico Vogt, Gregor Willkommen, Vladimir Schkolnik

Strongly Correlated Electrons

Kibble-Zurek Dynamics in Two-dimensional Frustrated Systems with a Neural Foundation-state Subspace Method

Universal scaling generated when a strongly interacting quantum many-body system is driven across a continuous phase transition provides a dynamical probe of equilibrium criticality. Accessing this regime numerically in two dimensions is challenging because it requires accurate real-time evolution of correlated many-body states over many system sizes and driving rates. We introduce a Neural Foundation-state Subspace (NFS) method for near-adiabatic dynamics. A foundation neural-network quantum state represents the ground-state manifold along the driving path, and a small fidelity-selected subset defines a fixed variational subspace. The many-body Schrödinger equation then reduces to the evolution of a few linear coefficients, with projected operators reusable across ramp times. We validate the method on the two-dimensional transverse-field Ising model, recovering the expected Kibble-Zurek scaling and critical exponents in quantitative agreement with ground-state quantum Monte Carlo estimates. Applied to the frustrated square-lattice $J_1$-$J_2$ Heisenberg model up to $16 \times 16$ clusters, our approach provides strong numerical evidence of Kibble-Zurek mechanism across the Néel-to-spin-liquid transition at $J_2/J_1=0.49$, yielding ${\nu=1.23(15)}$ and ${\eta=0.409(19)}$ at fixed $z=1$, consistent with static estimates and supporting the proposed continuous critical behavior.
Read the full abstract

All authors Linda Mauron, Luciano Loris Viteritti, Zakari Denis, Riccardo Rende, Giuseppe Carleo

Quantum Physics

The I3322 quantum value is attained spatially but not in finite dimension

Let $S$ be the common tensor-product and commuting-operator supremum of the $I_{3322}$ Bell functional in the Collins-Gisin normalization. Starting from a certified value window and Bellman/path equivalence, we prove that no finite-dimensional quantum strategy attains $S$, whereas a spatial strategy on $\ell^2(\mathbb{Z})\otimes\ell^2(\mathbb{Z})$ does. The finite-dimensional statement includes mixed states and binary POVMs. Consequently $C_q(3,3;2,2)$ is not closed and $C_{qs}(3,3;2,2)\setminus C_q(3,3;2,2)$ is nonempty. We also establish the dimension complexity $D(\epsilon)=\Theta(\log(1/\epsilon))$: approaching $S$ requires and suffices local dimension logarithmic in inverse error. The constructive upper bound is $D(\epsilon)\le 23.9010650\log(1/\epsilon)$ for all sufficiently small $\epsilon$, with natural logarithms; the lower constants are existential. The proofs use critical Bellman storage, spectral transport and normalizable orbit measures. A finite weighted-flow argument supplies the unrestricted quantitative lower bound without identifying distinct joint spectral components. This revision replaces an unsupported step in the earlier lower argument and records additional proof corrections. Prior numerical certification and independent concurrent attainment results are credited. Exact arithmetic and Lean 4 check specified numerical, scalar and finite accounting facts; the complete analytic proof is not formalized.
Read the full abstract

2609.05555 · 3 Sep · 25 pages. v2: Replaced the lower-bound argument with a weighted-flow proof; main theorem statements unchanged. Proof and verification archive: https://doi.org/10.5281/zenodo.22698978. Verification guide and scoped machine checks: https://github.com/Apsiape/i3322-exact-wall

General Relativity and Quantum Cosmology

Quantum-gravity-inspired Alcubierre warp-drive geometries

String T-duality, through its correspondence with path-integral duality, endows the low-energy propagator with a zero-point length $l_0=2\pi\sqrt{\alpha'}$ that softens the short-distance behavior of gravitational fields. Motivated by the regular black-hole construction obtained from the corresponding smeared source, we formulate a T-duality-inspired Alcubierre geometry by identifying the warp profile with the complementary cumulative mass fraction of that source. For the effective one-scale choice $f(r_s)=1-r_s^3/(r_s^2+l^2)^{3/2}$, with $l^2=R^2+l_0^2$, the metric is $C^2$ at the bubble center and smooth elsewhere, while the energy density, its volume integral, and the York expansion are available in closed form. In particular, $E=-(15\pi/1024)v_s^2 l$ in geometric units and $|\rho_E|$ is bounded by a constant times $v_s^2/l^2$. Within this one-scale family, the model therefore removes the divergence associated with the profile-contraction limit $R\to 0$ at fixed velocity, because $l$ cannot fall below $l_0$. This should not be confused with a regularization of the conventional Alcubierre thin-wall limit at fixed macroscopic bubble radius, in which an independent wall thickness is taken to zero. We emphasize, however, that the construction is an effective ansatz rather than a derivation from string-corrected field equations: $R$ is a macroscopic profile scale, the choice $l^2=R^2+l_0^2$ is an interpolation, and the classical $l_0\to 0$ limit remains a smooth thick-walled profile rather than the distributional Alcubierre top hat. Exotic matter remains necessary, and neither semiclassical stability nor a modified quantum energy inequality is claimed without an explicit renormalized stress-tensor calculation.
Read the full abstract

Artificial Intelligence

EdgeMem: LLM-Free Agent Memory Construction and Retrieval via Evidence-Preserving Multi-Anchor Hypergraph

Agent memory allows LLM agents to use earlier interactions when answering new queries. Existing methods often compress interaction histories into summaries or other LLM-generated representations. Repeated generation adds cost and can discard answer-bearing details before the system knows what a future query will require. We propose EdgeMem, an agent-memory method built around a simple principle: preserve original interaction turns and organize them through complementary content, temporal, and episodic cues. EdgeMem realizes this principle with a multi-anchor hypergraph constructed by lightweight local processing. Retrieval directly returns source evidence and reserves LLM use for final answer generation, combining structured access to multi-session histories with faithful retention of the original conversation. Experiments on LoCoMo and LongMemEval-S show strong retrieval and memory-grounded question answering; on LoCoMo, EdgeMem achieves the highest strict-judge score among seven reproduced systems under a shared prompt (61.01 versus 58.70), while construction and retrieval require no generative-LLM calls. Overall, EdgeMem shows that preserving and organizing source evidence provides an effective and efficient foundation for agent memory without generative memory management.
Read the full abstract

All authors Zeyang Cui, Jiannong Cao, Zhiyuan Wen, Bo Yuan, Junlan Feng, Shengyuan Chen

Human-Computer Interaction

An Agent Model Abstraction for Human-AI Teaming Cognitive Coupling

Industrial environments increasingly rely on collaboration between humans and AI-enabled agents. Effective teamwork requires aligning how agents perceive situations, plan actions to pursue goals, and adapt to changing conditions, yet existing systems lack mechanisms for cross-agent cognitive processes coupling. This paper presents a conceptual cognitive agent model that formalises cognitive coupling through eight components: Input, Process, Output, State, Value, Memory, World Model, and Goal. The model abstracts how agents coordinate and co-regulate their cognitive cycles, providing a basis for analysing distributed cognition and designing cognitively interoperable human-AI systems.
Read the full abstract

All authors Kolitha Kottagaha W.M, Jos A.C. Bokhorst, Ben Gaffinet, Christos Emmanouilidis

History and Overview of Mathematics

A Mean Value Property in Powers of the Natural Square

Let $n\geq 1$ be an odd integer, set $m=(n-1)/2$ and let $M$ be an $n\times n$ matrix whose coefficients are of the form $M_{i,j}=aij+bi+cj+d$ where $0\leq i,j\leq n-1$. Then we prove that for all squares centered at the central coefficient $M_{m,m}$, the mean value of the square equals $M_{m,m}$.
Read the full abstract

Computer Vision

Subject-Relative Micro-Motion and Sleep Dynamics for Near-Infrared Video Sleep Staging

Near-infrared (NIR) video is a promising modality for contactless sleep monitoring, but recent video-based sleep staging methods often use it as a route to reconstructed respiratory/cardiac proxies or cross-modal physiological representations. We study video-only sleep staging under labels defined by polysomnography (PSG), where the model infers sleep stages from NIR video alone without explicit physiological proxy reconstruction or auxiliary physiological signal supervision. This tests whether NIR video itself can provide informative sleep-stage evidence, rather than only serving as an input for recovering physiological proxies. We propose ViNUSS (Video-Native Unmediated Sleep Staging), a framework that combines subject-relative micro-motion learning with full-night sleep dynamics modeling. Spatially anchored pre-spatial micro-motion encoding preserves localized temporal variation together with its spatial context. Within-subject stage contrast learns stage cues with respect to each subject's night-specific baseline. Two-scale sleep dynamics modeling captures within-epoch motion evolution and organizes epoch-level evidence into a coherent full-night sleep-stage trajectory. On 475 overnight NIR recordings (~3,250 hours), ViNUSS achieves 0.80 accuracy and 0.78 macro-F1 for four-class sleep staging. Interpretability analysis suggests attention to thoraco-abdominal periodic motion and gross body movements associated with arousals and position changes. These results support NIR video as an independently informative and complementary modality for PSG-defined sleep-stage estimation
Read the full abstract

All authors Kunmin Jang, You Rim Choi, Hun Heo, Heonjun Lee, Suahn Bae, Dongik Park, Hyun-Woo Shin, Hyung-Sin Kim

Classical Physics

Symmetry Promotion in Electromagnetism

A symmetric distribution of charges or currents typically produces an electromagnetic field having the same symmetry group. However, under special circumstances we show that a discrete symmetry of the source distribution leads to a larger continuous symmetry for the resulting field in restricted regions. The conditions for this promotion from discrete to continuous symmetry are that the fields be affine linear and the symmetry of the source distribution act irreducibly on the ambient space. We illustrate this effect with many examples. This phenomenon is pedagogically interesting both as a source of inspiration in electro- and magnetostatics and as an accessible example of group representation theory.
Read the full abstract

Astrophysics of Galaxies

A Spatio-Temporal Self-Propagating Log-Gaussian Cox-Hawkes Process for Star Formation Modelling

Stochastic self-propagating star formation models describe galactic structure through local triggering and feedback but are commonly formulated using discrete spatial cells and time steps. We extend the spatio-temporal log-Gaussian Cox-Hawkes framework to obtain the Spatio-Temporal Self-Propagating Log-Gaussian Cox-Hawkes Process, a continuous point process model for the locations and times of star-forming events. The model combines spontaneous formation, correlated environmental effects, outwardly propagating excitation, local inhibition, differential rotation, and saturation. An observation layer transforms the conditional event intensity into idealised maps of instantaneous star-forming arm emissivity and recent young stellar surface brightness. For the selected parameter setting, the displayed realisation exhibits transient flocculent spiral-like patterns without any deterministic spiral geometry being imposed. The Monte Carlo experiment finds similar event production with and without rotation, whereas the stationary-front case produces fewer events. The displayed maps further suggest that differential rotation contributes to the winding and spatial arrangement of activity. The proposed framework provides a basis for future statistical inference from spatially and temporally resolved observations of star formation.
Read the full abstract

Quantitative Methods

A Network-Based Biomarker of Morphological Disruption Associated with Breast Cancer Malignancy

Breast cancer diagnosis commonly considers individual nuclear morphological characteristics, whereas their joint organization is less frequently evaluated. We developed the Morphological Network Disruption Index (MNDI), a patient-level measure of morphological abnormality relative to a benign reference state. Using the Wisconsin Diagnostic Breast Cancer dataset, ten nuclear characteristics were represented as network nodes, with MNDI summarizing disruption across 45 pairwise feature configurations. Performance was evaluated using repeated stratified five-fold cross-validation. Malignant lesions exhibited substantially higher MNDI than benign lesions (mean: 3.454 vs. 1.165; p = 2.94e-69). MNDI achieved an AUC of 0.941 (95% CI: 0.918-0.961), with 85.9% sensitivity and 91.6% specificity. A classifier using network-derived descriptors achieved an AUC of AUC of 0.928 +/- 0.024. Independent evaluation in the BreaKHis histopathology cohort showed limited discrimination (AUC = 0.538, 95% CI: 0.401-0.668), indicating dependence on the underlying morphological representation. MNDI provides an interpretable framework for quantifying patient-specific morphological disruption, while individual implementations require validation within compatible feature spaces.
Read the full abstract