SEARCH · Search arXiv

Search Search arXiv

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

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,333 records · Page 74Linked to original sources

arXiv2026

Forward-Only Learning for Shared-Kernel Wavelet Neural Networks without Backpropagation

Backpropagation remains the dominant training paradigm, using backward gradients and local updates that can suffer from vanishing or exploding gradients in deep networks. Although the Forward-Forward algorithm avoids conventional backpropagation, it remains mainly local or layer-wise, motivating global whole-network learning using forward evaluations only. For shared-kernel wavelet neural networks, we therefore propose a CMA-ES-based global forward-only learning framework. We further introduce structure-aware parameter decomposition and channel-specific credit assignment, factorizing the search into three subspaces. On BSDS500, our method reduces test MSE from 4.1783 to 4.0130 and improves PSNR from 42.621 to 42.787~dB over Adam/BP without backward propagation. Beyond these gains, training requires only 6 training-set coverages, compared with 1500 epochs for BP. Ablations show that structural decomposition and aligned credit assignment are more important than increasing population size, demonstrating the feasibility of whole-network forward-only learning for compact reconstruction networks. These results suggest a potential route toward forward-only optimization of compact reconstruction models for resource-constrained imaging and edge or near-sensor systems.

cs.CV↗

arXiv2026

Moduli spaces of logarithmic connections associated with difference Painlevé equations of type $E^{(1)}_n$

In this paper, we provide an explicit correspondence between moduli spaces of parabolic logarithmic connections on the projective line with spectral type $(2^2,1^4,1^4)$ and $(3^2, 2^3, 1^6)$, and the spaces of initial conditions for difference Painlevé equations with symmetry of type $E^{(1)}_7$ and $E^{(1)}_8$, respectively. Furthermore, we realize the extended affine Weyl group symmetries of the space of initial conditions in terms of connections.

math.AG↗

arXiv2026

AirPlayground: Benchmarking Aerial Navigation across Diverse Reconstructed Environments

Large-scale aerial resources provide a foundation for learning navigation across diverse environments. However, existing resources remain limited in providing (1) densely sampled flight demonstrations for learning different policy formulations and (2) broad environment coverage for standardized closed-loop evaluation. To address these limitations, we introduce AirPlayground, a unified benchmark for training and evaluating aerial navigation policies across diverse environments. We first construct 94 photorealistic 3D Gaussian Splatting (3DGS) environments spanning indoor, urban outdoor, and vegetation settings, with aligned collision geometry. Building on these environments, we develop a scene-adaptive pipeline for automated expert flight collection, yielding 94K demonstrations and 46.5M synchronized observation-state-action samples at 50 Hz. We further provide an extensible API that supports training different policy formulations on this shared corpus and evaluating them under unified protocols in the corresponding environments. We benchmark seven baselines spanning four policy paradigms across three navigation settings in seen and unseen environments, and further examine scaling behavior with ACT. Scaling ACT's training corpus from 5 to 70 scenes raises median success rates on seen and unseen environments from 30.6% and 16.8% to 41.2% and 29.4%, respectively. Through this shared training and evaluation framework, AirPlayground supports analysis of how offline prediction errors relate to closed-loop outcomes. Our analyses highlight the roles of prediction errors, execution protocols, and scene interactions in closed-loop performance.

cs.RO↗

arXiv2026

S-PASS: Star-Topology Pinching-Antenna Systems with Pipelined Antenna Pre-Positioning

Pinching-antenna systems (PASS) flexibly reconfigure the wireless channel by positioning dielectric antennas along a waveguide near each user. Releasing and re-attaching an antenna, however, is a mechanical operation orders of magnitude slower than electronic switching, wasting much transmission time. Existing designs generally assume instant reconfiguration, neglect the mechanical latency, or keep the antenna fixed. This paper proposes S-PASS, a star architecture of waveguides sharing one radio-frequency chain through a switch, and pipelined antenna pre-positioning (PAP), a time-division protocol in which idle waveguides pre-position their antennas for their next users while one transmits. It derives in closed form the scheduling efficiency, the number of waveguides beyond which the latency costs no transmission time, and the latency up to which PAP outperforms fixed placement. Analytical and numerical results show that this latency grows with each added waveguide, and users at the outer ends of the waveguides gain most.

cs.IT↗

arXiv2026

Beyond Policy Engines: Binding-Complete Execution Admission

Policy engines determine whether evaluated policies permit a request, but a correct ALLOW decision does not establish complete governance-source participation, exact context binding, or sufficient grounds for execution admission. This paper examines the gap between policy evaluation and execution admission. We introduce the Binding-Complete Execution Admission Contract (BCEAC), a governance-relative contract for independently determined source obligations, exact policy and contextual binding, and typed admission outcomes under explicit trust assumptions. We propose a conditionally engine-agnostic architecture in which participating backends must satisfy declared semantic profiles, adapter contracts, and contextual correspondence obligations; an Execution Boundary Language (EBL) authoring syntax remains optional. The architecture separates governance resolution, policy evaluation, independent admission verification, and downstream execution mediation. We investigate two distinct experimental tracks: a finite Governance Source Completeness study and an OPA/Cedar experiment with archived native-engine records examined by offline review. The OPA/Cedar archive comprises 72 scenarios and 144 strict case-engine records (120 native-invoked and 24 pre-native), with eight additional permissive controls outside the strict denominator. The offline review did not rerun the original native OPA/Cedar executables. The results provide bounded evidence while retaining historical failures and unresolved implementation defects. The work does not establish constitutional adequacy, general semantic refinement, authenticated currentness, machine-checked execution-safety theorems, or protected external effects. Its contribution is an explicit, falsifiable framework for determining what must be established between a policy decision and an execution-admission judgment.

cs.CR↗

arXiv2026

CausGround: Causal Interaction Grounding for Adaptive Embodied Planning

Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.

cs.RO↗

arXiv2026

Choosing Spatial Resolution in the Residence-Time Approach to Local Diffusivity

Estimating position-dependent diffusivity with the residence-time approach (RTA) requires choosing an interval width that resolves spatial variation while probing time scales compatible with a local diffusion model. We develop a selection criterion that combines bulk diffusivity convergence with local model agreement measured by the Kullback-Leibler divergence between observed and predicted first-exit-time distributions, and assess global model agreement through independent propagator analysis against unbiased molecular dynamics. Applications to water in a fluid phospholipid bilayer and three permeants in a model skin lipid barrier show that local agreement alone does not ensure global fidelity: in the skin barrier, narrow intervals minimize exit-time mismatch but predict excessive spreading, whereas requiring bulk convergence selects wider intervals with better longer-time predictions. In propagator tests, RTA profiles at the selected interval widths outperform velocity-autocorrelation-based profiles and generally improve agreement relative to position-autocorrelation-based profiles in the skin barrier, while phospholipid-bilayer rankings depend on lag time and discrepancy metric; both autocorrelation methods use harmonically restrained simulations.

physics.chem-ph↗

arXiv2026

Monotonicity and evolution of total isoperimetric profiles under Minkowski regularization

We study the evolution of the total isoperimetric profile of a planar convex body under volume-preserving Minkowski regularization. Several scalar geometric quantities are known to be monotone along this flow; we prove that the total isoperimetric profile is monotone as well, and characterize the equality cases. We also establish an explicit reparametrization law, by which, for nonnegative times, the profile of the regularized body is determined by that of the initial one, and the order of profiles of bodies of equal area is preserved. As consequences, we obtain the uniform convergence of the profile to the one of the ball of equal area; its joint $\mathrm C^1$ regularity in the volume and the regularization parameter; the fact that it solves a nonlocal quasilinear Hamilton--Jacobi equation; and sharp inequalities for the profile.

math.AP↗

arXiv2026

Monte Carlo study of extending the BM@N tracking acceptance toward midrapidity with the Vertex Silicon Plane

The Vertex Silicon Plane (VSP), installed close to the target in BM@N Run 9, can extend the tracking acceptance of the central tracking system toward larger polar angles. Using Monte Carlo simulations of Xe+Sn collisions at 3 AGeV, we reconstruct short tracklets from the primary vertex, one VSP hit, and hits in the first two Gas Electron Multiplier (GEM) stations. These tracklets extend the reconstructed primary-proton coverage toward midrapidity and higher transverse momenta. For the 10-30% centrality interval of the independent validation dataset, the conditional reconstruction efficiency for reconstructible high-$p_T$ primary protons after BDT selection is 67.0%, while the tracklet purity reaches 86.5%. For $p_T>1.2$ GeV/c, the VSP-assisted tracklets increase the number of reconstructed primary protons by 17.5% relative to standard BM@N tracking. Of these additional protons, 84.6% lie within $|y|<0.3$, where the relative reconstruction gain reaches 45.5%. The added acceptance is concentrated in the high-$p_T$, near-midrapidity region relevant for identified-proton observables. Its use in data requires particle identification in the same region.

physics.ins-det↗

arXiv2026

Stellar Loci X: Selection of Nitrogen-Rich Giants Using GALEX NUV and Synthetic SDSS Photometry

Nitrogen-rich (N-rich) field stars (likely second-generation escapees from globular clusters) provide a powerful, but currently sample-limited, probe of globular-cluster disruption in the Galactic halo. We exploit the fact that enhanced NH, CN absorption depresses optical u band flux, while the near-ultraviolet (NUV) band is largely insensitive to nitrogen enhancement while remaining sensitive to overall metallicity. Combining GALEX NUV photometry with (synthetic) SDSS optical photometry, we construct a [Fe/H]-dependent stellar-locus model for giants and select significant redder u-g outliers as N-rich candidates. For a relaxed cut ([Fe/H] <= -0.8, 2 sigma) we identify 970 candidates, with an estimated completeness of 64-65% and purity of 41-62% from multiple validations. A stricter selection ([Fe/H] <= -1.0, 3 sigma) yields 220 candidates, with lower completeness (30-38%) but substantially higher purity (60-91%). The dominant contaminants show anomalous spectral energy distributions (e.g., UV excess), and are themselves useful targets for follow-up. The diverse kinematic and dynamical properties of our candidates suggest multiple origins. This scalable photometric approach enables systematic searches for N-rich halo stars and is well suited to forthcoming wide-field surveys such as CSST.

astro-ph.SR↗

arXiv2026

Probing New Physics with $b \to u \ell ν_{\ell}$ induced decays

We analyze $B^- \to μ^- \barν_μ$, $\bar{B}^0 \to π^+ μ^-\barν_μ$, and $B^- \to ρ^0 μ^- \barν_μ$ decays using the data on binned branching fractions recently measured by Belle II Collaboration for the semileptonic modes. We extend the low-energy Hamiltonian governing $b \to u \ell \barν_\ell$ transitions to include the full set of dimension-six, Lorentz-invariant operators with left-handed neutrinos, consistent with the Standard Model Effective Field Theory (SMEFT) framework. The Wilson coefficients are treated as complex, lepton-flavor-dependent effective couplings. Since the three selected channels are sensitive to different non-standard structures, we perform a combined analysis using Monte Carlo methods to constrain the New Physics couplings for three benchmark values of $|V_{ub}|$. We find that, while the Standard Model remains broadly compatible with the data, some of the couplings show a mild preference for non-zero values.

hep-ph↗

arXiv2026

Kinetic Fluctuation-Dissipation Relation in collisionless plasmas: a correlated noise source

We here focus on small-scale dynamics of collisionless plasmas by employing a generalized kinetic Langevin framework driven by a stochastic time-correlated forcing. Our investigation captures the steady-state Fluctuation-Dissipation Relation (FDR) where the energy injection at low frequencies is due to the time-correlated forcing and the small-scale energy conversion occurs via Landau damping, the latter --though formally reversible-- acting as an effective dissipation mechanism. We analytically extract the resulting asymptotic spectral scalings across multiple physical regimes and characterize the velocity-space cascade by decomposing the plasma distribution function with Hermite polynomials. The framework is then generalized to a multi-scale forcing scenario in which correlation rates are distributed according to a power-law distribution. This extension provides closed-form analytical expressions that may serve to interpret spacecraft in-situ observations. All analytical predictions are successfully validated through direct numerical integrations of the underlying kinetic equations.

physics.plasm-ph↗

arXiv2026

Property (R) with its stability for analytic functional system and under compact perturbations

Let $\mathcal{F}_T$ be the analytic functional system generated by a linear bounded operator $T$ on an infinite dimensional complex Hilbert space. This paper characterizes the stability of property (R), a variant of Weyl's theorem, for $\mathcal{F}_T$ and under compact perturbations in terms of the spectrum on consistency in Weyl. In particular, the stability of property (R) under compact perturbations is shown to imply the (R)-stability of $\mathcal{F}_T$.

math.SP↗

arXiv2026

\textit{Ab initio} no-core shell model calculations of the electric dipole polarizabilities and photoabsorption cross sections of the $A=3$ mirror nuclei

We present \textit{ab initio} no-core shell-model calculations of the electric dipole ($E1$) polarizabilities and photoabsorption cross sections of the $A=3$ mirror nuclei $^3$H and $^3$He using the Daejeon16 nucleon-nucleon interaction. The polarizabilities are obtained by direct summation over discretized continuum states in the harmonic-oscillator basis, while the same spectral information is used to reconstruct the cross sections. This work provides the first extension of the approach from the spin-zero $^4$He to nonzero-spin nuclei by calculating the $1/2^-$ and $3/2^-$ final-state contributions separately and demonstrating their model-space convergence. In both nuclei, the $3/2^-$ contribution is approximately twice the $1/2^-$ contribution. After quantifying finite-basis and excitation-energy-truncation uncertainties, we identify a clear difference between the mirror polarizabilities, induced entirely by the Coulomb interaction because the Daejeon16 interaction is isospin symmetric. The reconstructed cross sections remain stable under variations of the inversion parameters of the Lorentz integral transform and exhibit the same angular-momentum hierarchy. These results extend the discretized-continuum approach to integrated and energy-resolved $E1$ observables in nonzero-spin nuclei with multiple final-state angular momenta.

nucl-th↗

arXiv2026

On unit groups of purely infinite rings

We prove that the unit group of any purely infinite, regular, right self-injective ring admits a universal sequence of words (thus has strong distortion), is minimally almost periodic, has finite involution width, is verbally simple and verbally elliptic, and has uncountable uniform strong cofinality. These findings are consequences of more refined structural results, some of which provide concrete cardinal estimates involving the Goodearl--Boyle dimension of the ring. A central ingredient in our results is a comprehensive study of bounded normal generation, in particular with respect to certain classes of involutions. Along the way, we develop some abstract methods and concepts pertaining to geometric group theory.

math.GR↗

arXiv2026

Will AI kill us all---or just make mathematicians obsolete?

AI's recent advances have fueled predictions of human intellectual obsolescence and fears for the future of mathematics. I argue that, impressive as these achievements are, they do not establish AI's capacity to invent new concepts and theories or sustain autonomous recursive self-improvement. Human intelligence also involves judgment, imagination, and embodied experience that standard measures fail to capture. Mathematics extends beyond proving theorems: its creative core lies in conceptual discovery and understanding. AI can strengthen this enterprise, but realizing its promise requires protecting attribution, open exchange, confidentiality, and mathematical training. I propose an academic AI consortium to provide powerful research tools with enforceable guarantees safeguarding unpublished ideas.

math.HO↗

arXiv2026

Condensation and optimal targeting of repeated forcing in heterogeneous networks

Many networks receive repeated external forcing through a subset of entry nodes, for example infections imported from another population, information seeded by a campaign, or invasive species arriving through ports. We study how concentrating such forcing on important entry nodes, with seeding probability $π_i\propto w_i^γ$, changes the size of the resulting cascades in a stochastic SIR model on bipartite networks. The response to one seeding event is linear in the degree of the seeded node, so the relative gain from targeting does not depend on transmissibility. The targeting distribution is a Gibbs measure. For importance tails $p(w)\sim w^{-a}$ and responses growing as $w^η$, the targeted response diverges at $γ_1=a-1-η$ and the forcing condenses on a few nodes at $γ_2=a-1$. Under repeated forcing, the burden-maximizing exponent $γ^*$ is strictly positive, and in large populations depletion of nodes already reached bounds it by the condensation point, $γ^*\leγ_2$. The optimum decays as $γ^*\simeq K/x$ for intense forcing, where $x$ is the expected number of seeding events per entry node, and approaches the condensation point as $γ_2-γ^*\simeq(a-1)/\ln(1/x)$ for sparse forcing. A cut-off of the importance distribution rounds the transitions in populations smaller than a crossover size $n_c$ and removes them in larger ones. Networks calibrated to survey data on sexual partnerships have between $0.08\,n_c$ and $0.8\,n_c$ entry nodes, which places them in the finite-size regime, where the asymptotic results hold only qualitatively.

physics.soc-ph↗

arXiv2026

Experimental determination of the double $K$-shell vacancy creation probability in the electron-capture decay of $^{55}$Fe with a microcalorimeter

The probability of double $K$-shell vacancy production following electron-capture decay is determined for $^{55}$Fe using a high-resolution microcalorimeter. The $K$-hypersatellite lines arising from double $K$-shell-vacancy filling are directly resolved within the measured spectra of X-ray emissions which are dominated by Mn $K$-X-ray diagram lines. The energies of the Mn \textit{K}-hypersatellite lines, $Kα_2^h$, $Kα_1^h$, and $Kβ_{1,3}^h$, are measured as $(6143.5\pm0.5)$ eV, $(6162.1\pm1.5)$ eV and $(6811.9\pm2.6)$ eV, respectively. The relative intensities of \textit{K}-hypersatellite to diagram X-rays are measured to be $(1.30 \pm 0.07) \times 10^{-4}$. From the relative intensities together with established fluorescence yield values, the probability of double-vacancy creation per $K$-shell capture ($P_{KK}$) is determined as $(1.22 \pm 0.09) \times 10^{-4}$. Comparisons with previous experimental results and theoretical calculations are discussed.

nucl-ex↗