arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 1,243 records · Page 69Linked to original sources

arXiv2026

Beyond city pairs: Higher-order information in urban growth across borders

We apply higher-order information measures to urban growth across borders and spatial units. We analyse 8,808 functional urban areas using gridded GDP estimates allocated from national accounts. Negative estimated O-information occurs in 29.93% of cross-border triplets and 2.00% of domestic triplets. This fifteen-fold contrast falls to fourfold at comparable dependence strength and persists under data-quality restrictions. Target-specific decompositions show how information overlaps or combines across cities. Administrative subdivisions have mean O-information within matched random-map ranges inside countries. These findings show why interpreting geographic dependence requires specifying comparison cities and spatial units, while economic connections remain to be established independently.

physics.soc-ph↗

arXiv2026

Towards Field-Deployable Wearable IMU Systems for Occupational Activity Recognition

Wearable inertial measurement unit (IMU) systems for construction worker human activity recognition (HAR) routinely report aggregate accuracies above 95 %, rarely under genuine subject-independent (leave-one-subject-out (LOSO)) evaluation, masking the per-individual variance that determines real-world deployment performance. Using the VTT-ConIoT benchmark, we conduct, to our knowledge, the first systematic configuration sensitivity analysis of this dataset: six classifiers across five sensor placements and three window sizes under a single LOSO protocol, a personalised-calibration pilot, and a validated comparison of tabular GAN-based and standard time-series augmentation. Per-subject LOSO accuracy for the main multi-sensor, 16-class configuration ranges from 61.9 % to 88.7 % (mean 77.5 %), a 27 pp spread, though all data are laboratory-collected. Multi-sensor fusion and longer windows both improve accuracy; standard augmentation gives small, consistent gains, while Conditional Tabular GAN (CTGAN) augmentation actively harms accuracy with a demonstrated distributional-fidelity mechanism; and personalised calibration recovers 16.8 pp for the weakest participants from just 30 % of their own data. These findings provide evidence-grounded configuration guidance for scaling occupational body sensor networks towards field deployment

cs.LG↗

arXiv2026

Accurate ab initio calculations of the ground state properties of singly charged alkali metal hydrides

In this article, we investigate the ground state electronic properties of singly charged alkali metal hydrides using the single-reference coupled-cluster method with single and double excitations (CCSD), together with perturbative triple excitations (CCSD(T)). A large relativistic core-valence quintuple-zeta basis set has been employed. The calculated spectroscopic constants and molecular properties are compared with existing data in the literature. Further, the relevant vibrational parameters are determined by solving the vibrational Schrödinger equation utilizing potential energy curves and permanent dipole moment curves. Thereafter, spontaneous and black-body radiation induced transition rates are calculated to obtain the lifetimes of the vibrational states.

physics.atom-ph↗

arXiv2026

HiSTEVE 1.0: Harnessing Human-Interface VLA Agents for Open-World Games

Building virtual embodied agents that operate in open-ended games like human players is a long-standing goal requiring both long-horizon autonomy and reliable native-interface control. As a procedurally generated open world, Minecraft provides a natural testbed for studying the integration of these capabilities. Existing agents typically provide either native visuomotor control without long-horizon autonomy or high-level planning executed through environment APIs or predefined skills. To bridge this gap, we propose HiSTEVE, a hierarchical VLA agent with a capability-aware agent harness. At the orchestration level, we equip the harness with structured memory for tracking progress and a VLM planner for selecting subtasks within the VLA's capabilities. To reliably determine subtask completion, we further introduce a policy-coupled task completion state detector that conditions on VLA policy context to signal candidate completion. At the control level, a single VLA executes all subtasks with vision-only perception and native keyboard-and-mouse actions. To mitigate interference between heterogeneous GUI and non-GUI interactions, we introduce regime-specific RMSNorm branches and separate mouse-movement token embeddings. We also train the VLA through world knowledge injection, imitation learning, and reinforcement learning with environment feedback to improve reliability and recovery. HiSTEVE achieves state-of-the-art atomic-task performance, scoring 85.0% on MineStudio Bench and a 37.4% average success rate on OpenHA, while also completing long-horizon tasks such as crafting an iron pickaxe from scratch like human players.

cs.CV↗

arXiv2026

Prediction Uncertainty Controlled Computer Model Calibration

Computer-model calibration is usually undertaken in service of prediction, yet neither Bayesian Kennedy-O'Hagan (KOH) calibration nor amortized simulation-based inference is built to control the predictive error of the calibrated simulator. When parameters are weakly identified or the data-model discrepancy is misspecified, the simulator-output bands induced by a nominal posterior level can be far from nominal. We propose a two-layer framework that targets predictive miscoverage directly. First, a conditional flow-matching model learns an amortized posterior for the calibration parameters, trained on simulator outputs perturbed with Gaussian-process discrepancies of varying structure. Second, conformal risk control arranges posterior draws for the target observation into nested candidate sets and selects the smallest one whose fixed-rank pointwise bands attain a prescribed time-averaged miscoverage on a synthetic calibration sample. Under exchangeability of calibration and target observations, the bands control expected miscoverage in finite samples. In a synthetic benchmark, the proposed bands are roughly 20% the width of KOH-MCMC bands with miscoverage near the 10% target. In calibrating an age-structured SEIR model to the 2010 Malawi measles outbreak, they are again sharper and near-nominal, and the weeks they fail to cover coincide with the asymmetric epidemic decline, exposing structural inadequacy of the mechanistic model.

stat.ML↗

arXiv2026

A local-to-global principle for $f$-divergence homogenization

We study the effect of replacing a family of independent statistical experiments by independent repetitions of their average. We give a simple, complete characterization of the $f$-divergences for which this operation increases divergence by at most a constant depending only on the generator, uniformly in the number of observations and the underlying measurable space. The characterization is governed by a local-to-global principle: a comparison restricted to two observations on a binary alphabet already determines the validity of the general inequality. It admits an explicit scalar formulation, which simplifies further for symmetric divergences. The result applies to arbitrary convex generators and requires no differentiability or symmetry assumptions. It recovers the homogenization inequality for total variation and identifies exactly the admissible members of the power-divergence family. In particular, whenever uniform homogenization fails, the obstruction is already present in a family of binary experiments with two observations. Moreover, the global comparison constant is controlled uniformly over generators with a common bound in the local inequality.

math.PR↗

arXiv2026

Network topology shapes consensus and vulnerability in LLM populations

As large language models (LLMs) are deployed in interacting populations, their collective behaviour depends not only on the models but also on who interacts with whom. Studies of convention formation in LLM populations have mainly tested well-mixed interactions. We place policies extracted from four LLMs on fully connected, random regular, and heterogeneous networks, with $10^3$ to $10^5$ agents playing the naming game. Topology affects how quickly consensus emerges, sometimes whether it emerges at all, and, in rare cases, which convention wins. Sparsest networks slow consensus up to twelvefold. Yet, by preserving correlations between persistent neighbours, they restore consensus in some cases that fail to converge under well-mixed interactions. Networks also govern vulnerability to adversarial attack. We measure the critical mass of committed agents needed to overturn an established convention. Tipping thresholds follow the same ordering across models and conditions: hubs $\leq$ random $\leq$ low-degree nodes. On the most heterogeneous topology, median critical masses are 3%, 16%, and 35%, respectively; targeting hubs requires one-sixth as many committed agents as random placement. For most models, networked populations are easier to tip than well-mixed ones, so well-mixed estimates can overstate robustness. Assessing the safety of multi-agent AI requires mapping who interacts with whom.

physics.soc-ph↗

arXiv2026

Memory before Goal: Role of Pre-History Memory for Humanoid Loco-Manipulation

Humanoid loco-manipulation in dynamic scenes is non-Markovian: cues vanish as objects change or are hidden, or leave the view as the robot turns and walks, yet whole-body vision-language-action (VLA) policies see only the present. We present WholeBodyMem, an object-guided perceptual memory that keeps the pre-history of object events seen before the robot acts on its goal. SAM3 tracks the task objects, and their events causally select key frames within an adaptive budget, which the action expert reads by cross-attention together with object tokens encoding appearance, geometry and time. We also contribute HuMeGo, a humanoid memory benchmark with six simulated and six real-world Unitree G1 tasks. Keeping the right frames rather than more frames, WholeBodyMem surpasses memoryless and uniform-memory baselines, raising its base policy's success from 29.1% to 72.6% in simulation and from 36.7% to 88.3% on the real robot, while also improving the standard SIMPLE tasks (70.2% to 86.9%).

cs.RO↗

arXiv2026

Liouville Theorems and Sharp Homogeneity Defects for Weighted Potential Systems

In this paper, we consider \[ \begin{cases} u=K*(|x|^αa_0(x)v^p)&\quad\hbox{in }\mathbb R^n,\\ v=K*(|x|^βb_0(x)u^q)&\quad\hbox{in }\mathbb R^n, \end{cases} \] where $n\ge2$, $0<ρ 0$, and $α,β>-ρ$. The radial kernel is nonincreasing and satisfies $K(r)\asymp r^{ρ-n}$. The coefficients satisfy $0 n-ρ+\varepsilon_K, \] where the defects bound the radial logarithmic derivatives. No symmetry, finite-energy condition, or prescribed decay is assumed. The proof combines a localized virial identity with a source-layer pressure estimate whose constants are uniform under centered dilations, including the low-order range $0<ρ\le1$. We also identify the necessary scaling range for existence. For radial weighted Newton systems with unequal powers, we prove that the combined coefficient defect has a sharp, unattained infimum equal to the subcritical gap. A fixed bound on the coefficient ratios forces a strictly larger defect, whereas suitable coefficient--solution pairs approach the threshold with excess at most $C/\log M$. In the scalar Newton case, the threshold is explicit and the approximating family converges, after normalization, to an explicit critical weighted profile. The sharpness results concern the Newton kernel; no corresponding construction is asserted for general fractional kernels.

math.AP↗

arXiv2026

SARIA: Persistence-Aware KV Caching with Coordinated DRAM--SSD Reload for LLM Agents

Multi-turn LLM agents usually reuse long context prefixes, making key value (KV) caching crucial for reducing prefill latency. SSDs provide a low-cost extension to limited DRAM capacity but introduce two key challenges. Eagerly persisting all KV cache blocks incurs substantial write traffic and rapidly consumes finite SSD endurance. Meanwhile, the small number of SSDs provisioned in typical GPU servers limits aggregate reload bandwidth, making SSD-resident KV cache much slower to retrieve than that in DRAM and inflating prefill latency. We present \textsc{Saria} (Selective Admission and Reload for Inference Acceleration), a KV caching system that jointly manages DRAM and SSD by exploiting the session-level temporal locality of LLM agents to reduce SSD writes and accelerate cache reload. On the write path, \textsc{Saria} uses eviction and reuse history to selectively persist KV cache blocks that exhibit reuse opportunities beyond their DRAM residency, avoiding unnecessary writes for KV cache already hosted in DRAM. On the read path, \textsc{Saria} promotes selected SSD-hit KV cache blocks from GPU memory into DRAM after prefill, allowing subsequent reuse within the same session to reduce SSD read traffic while limiting DRAM contention. A transfer coordinator paces DRAM reads while SSD transfers are active, allowing both paths to make progress and efficiently share GPU ingress bandwidth. We implement \textsc{Saria} on LMCache MP and evaluate it using workloads from TraceLab and WEKA, covering both full-attention and hybrid-attention models. The results show that \textsc{Saria} reduces SSD writes by 57.1--58.9\% compared with popular full duplicate caching and the mean time to first token (TTFT) by 42.3--79.7\% over DRAM-only caching.

cs.OS↗

arXiv2026

Quantum Constrained Hamiltonian Descent

We propose Quantum Constrained Hamiltonian Descent (QCHD), a continuous-time quantum framework for optimization under smooth equality constraints. The unconstrained counterpart, Quantum Hamiltonian Descent (QHD), minimizes a cost function by simulating the Schrödinger dynamics of a particle in the potential given by the cost function. Applying Dirac's formalism to the classical dynamics with constraints, we construct an intrinsic quantum Hamiltonian with a Laplace--Beltrami operator on the constraint manifold in a kinetic term. Its evolution preserves the constraints exactly without penalty terms and is expected to solve quantum-dynamically the constrained optimization problem. We analyze its relation to Quantum Hamiltonian Descent supplemented by a strong penalty term. For a squared-distance penalty, the effective tangential dynamics acquires an extrinsic-curvature potential and, in nontrivial normal-rotation sectors, a coupling to the normal-bundle connection. These embedding-dependent terms remain in the strong-penalty limit, generally yielding quantum dynamics different from QCHD. Numerical comparisons for circular, elliptical, and two simultaneous constraints show faster optimization and improved concentration near global minima for the initial states and finite penalty strengths considered.

quant-ph↗

arXiv2026

Optimal Hardy weights for the fractional discrete Dirichlet Laplacian on the half-line

For positive powers of the discrete Dirichlet Laplacian $(-Δ_{\mathrm{D}})^σ$, non-trivial Hardy weights exist if and only if $0<σ<3/2$. For $0<σ\leq 1$, an explicit optimal Hardy weight for $(-Δ_{\mathrm{D}})^σ$ was obtained by Das and de la Fuente-Fernández. Their ground state argument does not apply when $1<σ<3/2$, because the off-diagonal matrix entries of $(-Δ_{\mathrm{D}})^σ$ are no longer all non-positive. Nevertheless, we show that their result is true for the entire subcriticality range $0<σ<3/2$. For $1<σ<3/2$, our argument combines the factorisation $$(-Δ_{\mathrm{D}})^σ=\nabla_{\mathrm{D}}^{*}(-Δ_{\mathrm{N}})^{σ-1}\nabla_{\mathrm{D}},$$ where $\nabla_{\mathrm{D}}$ and $-Δ_{\mathrm{N}}$ are the discrete Dirichlet gradient and the discrete Neumann Laplacian, respectively, with a suitable two-term ground state representation. In addition, we use these arguments to derive another explicit family of optimal Hardy weights for any $σ\in(0,3/2)$.

math.SP↗

arXiv2026

Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness

LLM-as-a-Judge evaluation is increasingly applied to multilingual tasks, yet prompting strategies such as Chain-of-Thought and Role Conditioning are recommended based solely on English benchmarks. Whether their effectiveness transfers across judge models and languages remains unknown: prompt-sensitivity studies are largely English-centric, while existing multilingual judge studies fix the prompt template. In this work, we address this gap by measuring this interaction through a benchmark of 9 prompting strategies, 4 open-weight judges, 27 languages, and 4 tasks across 13 datasets, yielding over 1M pointwise scores on gold-reference outputs, validated against human judgments. We find that prompting strategies are not transferable: the choice of the judge model dominates the choice of prompting strategy, strategy effects reverse direction across judges, and stability degrades in low-resource languages, where these effects are concealed by aggregate reporting. These findings show that prompting effects in multilingual evaluation are conditioned on the judge, task, and language. Consequently, prompting strategies should be validated for the intended evaluation setting, and cross-study comparisons should account for both judge and prompt configuration.

cs.CL↗

arXiv2026

On the arithmetic Siegel--Weil formula for orthogonal Shimura varieties over a general totally real field

Let $\mathrm{F}/\mathbb{Q}_p$ be a finite extension with $p\ne 2$, and let $V$ be a quadratic space over $\mathrm{F}$ of dimension at least two admitting a self-dual lattice. We prove the local arithmetic Siegel-Weil formula for GSpin Rapoport-Zink spaces formulated by Li and Zhang. We use the global-to-local strategy of Mihatsch and Zhang and its application by Luo. Using a generalization of the prismatic realization functor of Imai, Kato, and Youcis, we construct and study derived special cycles on integral models of SO Shimura varieties and their Rapoport-Zink uniformizations. We derive the local intersection formula from the modularity of special cycles on the generic fiber. As a first application, we prove the corresponding semi-global arithmetic Siegel-Weil identity for orthogonal Shimura varieties over a totally real field $F\ne\mathbb{Q}$.

math.NT↗

arXiv2026

An elementary proof of rational injectivity in equivariant KK-theory

Antonini, Azzali and Skandalis proved that, for every countable discrete group $Γ$, the canonical map $EΓ\to\underline{E}Γ$ induces a split injection in equivariant $KK$-theory with real coefficients. In this paper, we give a more elementary proof of the corresponding rational injectivity result in ordinary equivariant $KK$-theory without using $\mathrm{II}_1$ factors.

math.KT↗

arXiv2026

Robust paths to primordial $B$ modes in the presence of polarised point source contamination

A detection of primordial $B$ modes in the Cosmic Microwave Background (CMB) polarisation relies on a robust control and removal of contamination from astrophysical polarised foreground emission. We examine how much polarised extragalactic point sources may contaminate the cosmological signal in ground-based CMB experiments, propagating their impact through component separation to the final parameter constraints. We discuss masking and modelling strategies to ensure that the resulting constraints remain both accurate and precise under this source of contamination. We produce synthetic data containing polarised emission from extragalactic radio sources (ERSs) and dusty star-forming galaxies (DSFGs) following state-of-the-art models for both samples. We find that, while the DSFG contribution is likely too small to cause any significant bias to the cosmological or foreground parameters, the ERS contribution instead, if left unmasked, leads to a bias in the parameters describing the Galactic synchrotron and lensing $B$-mode components. For experiments targeting noise levels of order $σ_P\simeq 1\,μK\,{\rm arcmin}$, we find that masking all ERSs with fluxes above $\sim$0.1 Jy, coupled with a relatively simple model to marginalise over residual point-source contamination, achieves a suitable balance that allows for the robust recovery of unbiased cosmological and foreground parameters at a negligible cost in parameter uncertainties.

astro-ph.CO↗

arXiv2026

Tracking the Emergence and Persistence of Abnormal Grain Growth in an Aluminum Alloy using Diffraction-Contrast Tomography

Abnormal grain growth is inherently three-dimensional, making 3D tracking of grain evolution essential for understanding its underlying mechanisms. This study combines synchrotron-based diffraction-contrast tomography and phase-contrast tomography to characterize the three-dimensional grain structure and quantify the size and spatial distribution of second-phase particles in the aluminum alloy AA5252. Using an interrupted annealing protocol, we tracked the persistent growth of abnormal grains and the emergence of abnormal growth in a grain initially belonging to the matrix. Although second-phase particles effectively suppressed grain growth in the matrix, variations in local particle density showed no apparent correlation with either the emergence or subsequent rapid growth of abnormal grains. Instead, crystallographic misorientation and, more notably, neighboring grain size were identified as stronger predictors of abnormal growth rate. Matrix grains were consumed more rapidly when they were relatively small and shared high-angle grain boundaries with the growing abnormal grain. The same trends were observed for the grain that underwent a transition from normal to abnormal growth. Remarkably, even before the onset of abnormal growth, this grain was surrounded by smaller matrix grains with exclusively high-angle grain boundaries, while the local particle density was slightly higher than average. These findings suggest that the same microstructural characteristics facilitate both the emergence and persistence of abnormal grain growth, although the mechanism triggering the transition to rapid growth remains unresolved.

cond-mat.mtrl-sci↗

arXiv2026

Gravitational waves from highly charged binary black holes in Einstein-Maxwell-dilaton theory

We perform fully nonlinear numerical-relativity simulations of electrically charged binary black hole mergers in Einstein-Maxwell-dilaton (EMD) theory, varying the magnitude and relative sign of the input charges and the dilaton coupling $a$. In the weak-charge sequence, the dominant gravitational waveforms remain close to the neutral vacuum baseline of general relativity and are visually indistinguishable across the sampled couplings, consistent with earlier EMD simulations. For the highly charged binaries with equal-magnitude opposite charges, however, the dominant gravitational strain exhibits visible differences between the EMD cases with $a\neq0$ and the charged Einstein-Maxwell reference at $a=0$. At $a=1$, the coordinate-time interval from the same separation to the waveform-based merger is approximately $62\%$ shorter than for the neutral binary. In this high-charge sequence at fixed input charges, increasing $a$ strengthens the scalar dressing while leaving the individual dilaton-weighted horizon charges nearly unchanged. Within the same sequence, increasing $a$ across the sampled values decreases the dipolar electromagnetic-flux peak, increases the scalar-flux peak, and modestly lengthens the merger interval. The sampled like-charged binaries have corresponding merger intervals approximately $15\%$-$20\%$ longer than the neutral baseline and emit predominantly quadrupolar electromagnetic radiation. At $a=1$ and $a=\sqrt{3}$, the like-charged mergers leave charged remnants with scalar fields consistent with secondary dilaton hair. The opposite-charge remnants instead approach neutral black holes without stationary scalar hair.

gr-qc↗