Statistical Computation
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Wednesday 9 September
We consider the problem of estimating a high-dimensional covariance matrix from a very limited number of samples. This problem is ubiquitous in computational fluid dynamics, where a small number of fluid snapshots must be used to construct a Gramian matrix determining a reduced-order model, as well as in computational geoscience, where a small ensemble of Earth system forecasts must be used to estimate the covariance matrix associated with the forecast uncertainty. It is common practice to regularize the small-sample covariance by imposing a "localization" structure that enforces a physically realistic correlation length scale, imposing a sparsity constraint, "shrinking" towards a prescribed target, or attenuating small correlations. We propose an alternate technique that regularizes the small-sample covariance by imposing hierarchical rank structure. Compared to regularization methods that assume sparsity such as spatial localization, hierarchical rank structure accommodates a wider range of covariance matrices, roughly corresponding to situations where long-range correlations vary more smoothly than short-range ones. It also results in a data-sparse matrix format that permits highly efficient matrix-vector products. We present theory and algorithms which show how to efficiently estimate a high-dimensional, hierarchically rank structured covariance matrix from limited samples. Through an error analysis and numerical experiments with a variety of model problems, we demonstrate that these techniques are effective at reducing sampling errors, and that in many cases they achieve smaller estimation error than conventional techniques.
Information Theory
Quantum MDS codes from complements of unions of finite-field subsets
Let $q$ be an odd prime power. We use complements of unions of subsets of $\mathbb F_{q^2}$ as locator sets and establish a sufficient condition under which a generalized Reed–Solomon (GRS) code is Hermitian self-orthogonal. Using cosets of multiplicative subgroups and sets with prescribed trace or norm values, we construct five families of Hermitian self-orthogonal GRS codes over $\mathbb F_{q^2}$. The Hermitian construction then yields five corresponding families of $q$-ary quantum maximum-distance-separable (MDS) codes. Under suitable parameter conditions, these quantum codes have minimum distances greater than $q/2+1$. By comparing codes of the same length, we give conditions under which our codes have strictly larger minimum distances than those obtainable from several previously known constructions based on trace maps, linear transformations, and cosets of multiplicative subgroups, either directly or via the propagation rule. We further show that such improvements occur for infinitely many values of $q$.
Read the full abstract
Number Theory
Non-vanishing of symmetric square $L$-functions in the weight aspect
We prove a new asymptotic formula for the second moment of symmetric square L-functions in the weight aspect on average. This result implies that the associated L-function is non-vanishing at the central point for at least 76.69% of holomorphic Hecke cusp forms of bounded weight, improving the previous bound of 70.37%.
Read the full abstract
Robotics
HaWMPO: Hallucination-Aware World Model-based Policy Optimization for Generalist Robot Policy
Generalist robot policies have demonstrated strong generalization across robotic manipulation tasks, yet their success rates remain limited in com- plex long-horizon scenarios. Recent methods improve Visual-Language-Action (VLA) policies through online reinforcement learning on real robots, but such training relies on costly physical interactions, suffers from low sample efficiency, and may introduce hardware and safety risks. World models offer a promising alternative by enabling policy optimization with imagined rollouts. However, long-horizon rollouts generated by world models often suffer from prediction hal- lucinations, producing biased state transitions that can mislead policy learning. To address this issue, we propose Hallucination-aware World Model-based Pol- icy Optimization (HaWMPO), a closed-loop reinforcement learning pipeline for VLA policy post-training with world models. Specifically, HaWMPO introduces an action-conditioned hallucination-aware model to estimate the reliability of gen- erated image sequences, and incorporates hallucination scores into group relative policy optimization through a Reward-Soft mechanism, suppressing unreliable ac- tion chunks during training. On the LIBERO benchmark, HaWMPO achieves the best average success rate, with gains of 15.0% over the base model and 2.8% over the strongest baseline; real-world experiments on a G1 robot further validate its effectiveness, raising the average success rate on two manipulation tasks from 67.5% to 80.0%.
Read the full abstract
Audio and Speech Processing
NVV-Locator: From Transcript Tags to Acoustic Boundaries for Fine-Grained Nonverbal Vocalization Grounding
Human speech includes nonverbal vocalizations (NVVs), such as laughter, sighs, breaths, and coughs, which convey affective and interactional information. Existing approaches typically represent NVVs as transcript-level tags, providing limited supervision for their waveform-time boundaries. We present NVV-Locator for fine-grained NVV temporal grounding. We first unify 26 NVV categories across public resources and construct large-scale timestamp-supervised training data through dual-LLM verification, transcript-guided forced alignment, and energy-based boundary refinement. We further introduce NVV-TimeBench, an expert-refined benchmark with 667 utterances and 1,094 events. NVV-Locator uses a non-autoregressive slot-filling architecture to jointly predict lexical timestamps, NVV categories, and event boundaries. On NVV-TimeBench, it achieves 71.0% Micro F1, 70.2% Macro F1, 80.4% Macro mIoU, and 59.6 ms Macro mMAE, outperforming the evaluated large audio model counterparts. Evaluation on an external corpus further demonstrates the cross-corpus generalization of NVV-Locator.
Read the full abstract
Optics
Image recognition based on optical spike processing with exciton-polaritons
Exciton-polariton condensates combine ultrafast dynamics with strong optical nonlinearities, making them promising for optical neuromorphic computing. Here, we numerically demonstrate image recognition using the transient response of an exciton-polariton condensate. Downscaled MNIST images are encoded as optical pulse sequences, and the resulting condensate dynamics are sampled and processed by a linear classifier. For ten-class classification, condensate features achieve 92.2% accuracy, compared with 91.4% for raw-pixel linear classification. Combining both feature sets increases the accuracy to 93.2%. For the challenging binary classification of digits 7 and 9, the combined model reaches 97.3%, exceeding both the linear (95.5%) and feed-forward neural-network (96.6%) baselines. These results demonstrate that polariton dynamics can extract useful nonlinear features for ultrafast optical information processing.
Read the full abstract
Functional Analysis
A Dimension-Independent Commutator Bound
We prove that every trace-zero matrix $A\in M_n(\mathbb{C})$ admits a representation $A=BC-CB$ with $B,C\in M_n(\mathbb{C})$ and $\lVert B\rVert\lVert C\rVert\le K\lVert A\rVert$, where $K$ is an absolute constant independent of $n$, and $\lVert\cdot\rVert$ denotes the operator norm. For a fixed $t>0$, the proof splits according to whether $\lVert\operatorname{Re}(e^{\mathrm{i}\theta}A)\rVert_1\ge tn\lVert A\rVert$ holds for all $\theta\in\mathbb{R}$, where $\lVert\cdot\rVert_1$ denotes the trace norm. When this lower bound holds, we construct a commutator representation directly. Otherwise, the vector-selection theorem of Marcus, Spielman, and Srivastava yields smaller trace-zero compressions whose norms are small enough for the induction to close. We also construct an explicit family of zero-diagonal Hermitian unitaries that forces a lower bound of order $\sqrt{\log n}$ for $\lVert B\rVert\lVert C\rVert$ when either factor is required to be diagonal in the prescribed basis. The same family admits $\varepsilon$-pavings with fewer than $2\varepsilon^{-2}$ blocks and representations by two normal factors with optimal norm product $1/2$. This establishes a distinction between unrestricted commutator bounds and bounds under a prescribed diagonal restriction. The main results and their essential inputs are formalized in Lean 4 using Mathlib. The development also includes a formal derivation of the Kadison-Singer state-extension theorem from the same vector-selection theorem.
Read the full abstract
Geometric Topology
Links and singularities of stable maps from $3$-manifolds to surfaces
For any two links in the $3$-sphere, we provide a visual construction of a stable map $f$ from the $3$-sphere to the Euclidean plane such that $f$ has no cusp points, the set of definite (resp. indefinite) fold points of $f$ is isotopic to the first (resp. second) given link, and $f$ does not have a certain type of singular fibers. Moreover, we obtain a similar result for any $3$-manifold by setting the target space to the $2$-sphere.
Read the full abstract
Soft Condensed Matter
Configurational-space separation and structure selection in three hard squares
Self-assembly of hard particles with diverse shapes gives rise to a rich variety of structures through excluded-volume constraints alone. Here we show that even a minimal system of three hard squares confined in a two-dimensional periodic box exhibits nontrivial configurational behavior relevant to structure selection. As the packing fraction increases, radial distribution functions obtained from Markov-chain Monte Carlo and uniform non-overlapping insertion sampling agree at low densities, deviate markedly over an intermediate range, and converge again at higher densities. Pressure measurements provide strong numerical evidence that the discrepancy originates from the separation of the allowed configurational space into two disconnected regions above a characteristic density. We identify the separation density as $\phi_{\rm sep}=3/5$, construct explicit overlap-free transition pathways connecting the two regions immediately below it, and quantify their relative configurational-space volumes. At higher packing fractions, an approximately L-shaped arrangement of the particle centers becomes strongly favored over a staggered one, revealing a structural motif characteristic of tetratic and square-lattice ordering in larger hard-square systems. These results show that excluded-volume geometry can govern both configurational connectivity and local structure selection even in a three-particle system, revealing how signatures of many-particle self-assembly can already emerge in the few-particle limit.
Read the full abstract
Combinatorics
Oriented paths with two blocks in bipartite oriented graphs
Stein conjectured that for any integer $k\geq 2$, every oriented graph with minimum semidegree greater than $k/2$ contains every orientation of a path with $k$ edges. Recently, Chen, Hou and Zhou proved this conjecture to be true for any oriented path with two blocks, where a block of an oriented path is a maximal directed subpath within it. In this paper, we prove that every bipartite oriented graph with minimum semidegree at least $3k/8+2$ contains every oriented path with two blocks of length $k$ for $k\ge 2$. Moreover, in contrast to the general oriented setting, we highlight that the minimum semidegree threshold in the bipartite setting is closely related to the number of blocks.
Read the full abstract
Information Theory
Collision-Free Schedule Acquisition through Slotted ALOHA: Scaling Laws and Fluctuations
Decentralized networks can learn collision-free periodic transmission schedules, but acquisition can consume much of a finite operating horizon. We characterize acquisition in a reservation protocol for a fixed population of $n$ nodes and a period of $\Gamma_n\ge n$ slots. Initially, each node attempts independently in each slot with probability proportional to $1/n$. A success reserves a slot that the node reuses every $\Gamma_n$ slots, retaining it after collisions. Nodes use only their own acknowledgments and small local memory. An exact finite-population Markov reduction, independent of the arrangement of reserved slots, characterizes the time until every node has a distinct reservation. As $n\to\infty$, mean acquisition takes $O(n^2)$ slots when $\Gamma_n=n$ and $O(n\log n)$ slots when $\Gamma_n\sim(1+\theta)n$, for fixed $\theta>0$. With a fixed number $\Gamma_n-n$ of spare slots, the standard deviation of acquisition time has the same asymptotic order as its mean. If that number diverges while $\Gamma_n/n$ remains bounded, the ratio of standard deviation to mean tends to zero. A diverging sublinear number of spare slots yields $o(n^2)$ mean acquisition while long-run utilization, the fraction $n/\Gamma_n$ of slots with successful transmissions, tends to one. Simulations validate these laws and compare acquisition and delivered packets per slot with existing access protocols.
Read the full abstract
Solar and Stellar Astrophysics
Long-slit spectroscopy of the bipolar planetary nebula ESO\,428$-$05
The results of detailed long-slit spectroscopy of the bipolar planetary nebula \ESO\ are presented. The observations were obtained with the facility spectrograph RSS of the SALT telescope. The total measured extent of the nebula reaches $\approx$55$^{\prime\prime}$, and the kinematics along the slit directly confirms its bipolar structure. The inclination of the nebula to the line of sight is estimated as $i\approx 45^{\circ}$, and the kinematic ages of two morphological structures are measured: the expanding ring and the polar outflow in the direction perpendicular to the ring. The age of the former is $\approx$1.5$\times$10$^4$~yr and that of the latter is $\approx$(3–4)$\times$10$^4$~yr, i.e. the polar outflow is 2–3 times older than the central ring. The ratio of these ages does not depend on the adopted distance. The weighted mean heliocentric velocity of \ESO\ is V$_{\rm hel}$=74.7$\pm$1.6~\kms. Chemical abundances are obtained for 19 regions along the major axis of the nebula, over an extent of 40$^{\prime\prime}$, i.e. for different morphological parts of \ESO. The abundances were derived with the T$_{\rm e}$–method, where the electron temperature was calculated directly using the measured faint auroral lines [\ion{O}{iii}]~$\lambda$4363~Å\ and [\ion{N}{ii}]~$\lambda$5755~Å. The abundances of O, N, Ne, S, Ar, Cl and He were determined. Despite a significant stratification, the measured abundances along the slit can be considered nearly constant.
Read the full abstract
Numerical Analysis
Order Reduction of Exponential Runge–Kutta Methods: Fourth-Order Schemes for Non-Commuting Operators
This paper extends the convergence analysis of explicit exponential Runge–Kutta methods for linear parabolic problems $u'(t) + Au(t) = Bu(t)$, where $A$ generates an analytic semigroup and $B$ is relatively bounded with respect to $A$, from the third-order case to fourth-order schemes. By establishing the global error recursion relation and extending the defect-based analytical framework, we identify the terms responsible for stiff order reduction when $A$ and $B$ do not commute. Numerical experiments are performed on a non-commuting advection-diffusion problem to validate the theoretical results. Numerical tests using the classical four-stage, fourth-order schemes of Krogstad and Strehmel & Weiner exhibit an observed convergence order of approximately 2.75, which matches the theoretical prediction from the convergence analysis.
Read the full abstract
General Relativity and Quantum Cosmology
Revealing the Quantum Signature of Gravity via Gravitational Waves
Can propagating gravitational waves serve as operational probes of the quantum nature of gravity? We address this question by developing a unified theoretical framework that combines spacetime geometry, quantum information, and gravitational-wave physics. Starting from the geodesic deviation equation in linearized General Relativity, we derive the effective detector Hamiltonian directly from spacetime geometry and construct the complete quantum dynamics for detector subsystems interacting with both classical and quantized propagating gravitational-wave fields. This unified formulation enables a direct comparison between classical and quantum descriptions of gravitational radiation within the same physical framework. We demonstrate that classical gravitational-wave backgrounds can induce mixedness in the detector state but cannot generate genuine quantum correlations between the detector subsystems. In contrast, quantized gravitational waves coherently mediate gravity-induced entanglement, quantum coherence, quantum memory, and nonclassical correlations, providing clear operational signatures of the quantum nature of propagating gravitational radiation. We further discuss how mesoscopic quantum mechanical oscillators offer a promising route towards experimentally probing these effects. Our results establish a geometric and quantum-information-based framework for exploring quantum gravity through propagating gravitational waves.
Read the full abstract
Optimization and Control
The Global Asymptotic Stability Problem for Linear MPC Is Undecidable
We prove that deciding global asymptotic stability for constrained finite-horizon linear model predictive control is undecidable. This holds at horizon one with identity state, input, and terminal weights, unique optimizers, and global feasibility. Separate reductions cover predicted-state boxes, hard input boxes, and quadratically softened input boxes. A fourth reduction fixes the state and input dimensions to three and six. Hence undecidability is not caused by long horizons, growing dimensions, failures of recursive feasibility, or nonuniqueness.
Read the full abstract
Audio and Speech Processing
Source-Adaptive Data Curation for Bilingual NVV-Aware ASR
Nonverbal vocalizations (NVVs), such as laughter, sighs, breaths, and coughs, convey affective and interactional information that conventional automatic speech recognition (ASR) systems often discard. We present a bilingual Mandarin-English system for Track 1 of the NVVSpeech Challenge at ISCSLP 2026, which requires joint transcription of lexical content and 16 NVV categories at their transcript-relative positions. Our NVV-Aware Whisper adapts Whisper-medium through checkpoint-compatible vocabulary remapping, enabling lexical tokens and inline NVV tags to be decoded within a unified autoregressive sequence without expanding the vocabulary. To provide reliable and diverse supervision, we further introduce a source-adaptive data curation strategy that refines public NVV corpora through acoustic augmentation and multimodal LLM filtering, while mining spontaneous NVVs from in-the-wild media through automated preprocessing and annotation. Under the official bilingual evaluation protocol, the proposed system improves final score from 33.32 to 53.61, with ablations confirming the complementary benefits of the proposed data-curation components.
Read the full abstract
Artificial Intelligence
Structural Process Supervision for Latent Chain-of-Thought Reasoning
Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings. However, existing methods lack efficient process supervision over these embeddings, which often leads to representation collapse and uneven information distribution. % However, supervising these latent embeddings is challenging because a fixed number of latent states must capture information from variable-length reasoning traces. To address this, we propose Prototype-Mediated Process Supervision (PMPS), which introduces learnable reasoning prototypes as semantic anchors to provide structural process-level supervision for latent reasoning. PMPS projects latent embeddings and explicit CoT embeddings into a shared prototype space, achieving many-to-many soft alignment between unequal-length representations through prototype assignment. Meanwhile, we introduce a Progressive Sequential Alignment (PSA) module to further guide training: positional priors initially encourage sequential alignment structure, then gradually relax to permit adaptive matching. Experimental results show that PMPS compresses output token length to under 50% of explicit CoT on GSM8K-Aug. Compared to leading baseline SIM-CoT, our method achieves average accuracy gains of 2.51% across different model families. On GPT-2, accuracy of PMPS even surpasses CoT-SFT. On larger models and a more challenging task, PMPS consistently attains the highest accuracy among all latent reasoning methods with comparable output length.
Read the full abstract
History and Philosophy of Physics
Is Black Hole Evaporation Prediction Friendly?
Manchak and Weatherall (2018) formulate the black hole information paradox as a failure of predictability in black hole evaporation spacetimes, diagnosed by non-global hyperbolicity. I offer a strategy for resolving this paradox. I argue that failures of predictability in black hole evaporation are not well diagnosed by non-global hyperbolicity. I then consider two weakenings of global hyperbolicity: prediction and retrodiction friendliness, the failure of which could ground a new paradox. However, deidealized black hole evaporation models can be prediction and retrodiction friendly. Therefore, the information paradox cannot be based upon failures of global hyperbolicity, nor either retrodiction or prediction unfriendliness.
Read the full abstract
General Relativity and Quantum Cosmology
Prospects of electromagnetic follow-up of neutron star-black hole mergers in the LIGO-India era
Neutron star-black hole (NSBH) mergers are promising multimessenger sources, but only a subset is expected to produce detectable electromagnetic (EM) counterparts, depending on the binary mass ratio, black hole (BH) spin, and neutron star (NS) equation of state (EoS). We investigate the prospects for detecting kilonova counterparts to NSBH mergers in the LIGO-India era using the Vera C. Rubin Observatory (Rubin). We simulate GW230529-motivated NSBH populations with two effective tidal deformability ranges, estimate gravitational-wave detections for the LHV (LIGO-Livingston, LIGO-Hanford, Virgo) and LHVA (LHV with LIGO-India at Aundha) detector networks, and model the detectability of the associated kilonova emission with Rubin. We find that adding LIGO-India approximately doubles the number of EM follow-ups, primarily by increasing the duty cycle and reducing the median sky-localization area, and notably increases the number of such detections at large distances. For fixed-exposure strategies, the EM detection rate saturates at exposure times of a few hundred seconds, reflecting the trade-off between depth and sky coverage. We therefore introduce an event-specific exposure-time optimization strategy based on the expected counterpart brightness and sky-localization area. With this strategy, the expected NSBH GW+EM detection rate per year is 4.19(-3.58,+9.66) for the optimistic and 0.79(-0.68,+1.83) for the conservative population with the LHVA network, compared to 1.46(-1.25,+3.37) and 0.27(-0.23,+0.62), respectively, for LHV. The required Rubin follow-up time remains within the expected Target-of-Opportunity allocation. These results show that in the next decade, a few such NSBH follow-ups may take place, with a fraction of them in the redshift range of about 0.1-0.2, which will not only be useful for measuring the Hubble constant but can also help us in probing the Hubble parameter.
Read the full abstract
Artificial Intelligence
Time-Frequency Geometric Cross-Attention for Chunked Vision-Language-Action Models
Modern vision-language-action (VLA) policies predict a whole chunk of actions: one to two seconds of coordinated motion emitted in a single forward pass. Yet an action chunk is essentially a short multivariate trajectory, but inside these models it is a sequence of generic per-timestep hidden tokens decoded by a linear head. This under-serves two motion structures. First, frequency: a chunk superimposes a smooth global trend and fine corrective motion across time scales, and a single token entangles them. Second, cross-phase geometry: motions of different phases (reach, contact, grasp adjustment, settling) unfold along very different, near-orthogonal directions in representation space, yet are tightly related for the task and arise across the time axis. Dot-product attention scores alignment by an inner product, so it favors aligned tokens and is least sensitive near orthogonality, leaving such relationships for the network to recover through a detour. We introduce Time-Frequency Geometric Cross-Attention (TFGCA), a drop-in module repairing both blind spots. TFGCA uses a per-dimension learnable stationary wavelet transform to decompose the action chunk into time-frequency tokens, and each time token retrieves information from them via a cross-attention that fuses the dot product (similarity) with the wedge-product magnitude (sensitive to near-orthogonality) through a learnable weight. A zero-initialized residual reproduces the base behavior at initialization, so it can be dropped onto a pretrained VLA and fine-tuned jointly. Relative to the same-source base, TFGCA improves in-distribution LIBERO by +1.5 on average, the OOD LIBERO-Plus by +6.3, the randomized average under RoboTwin domain randomization by +28.5, and the overall success rate on three real-robot AgiBot A2 tasks by +11.67 points, with larger gains out of distribution.
Read the full abstract
Computer Vision
Multimodal Emotion Recognition in Conversations via Class-Wise Adaptive Modality Fusion and Affective Geometry
Emotion Recognition in Conversations (ERC) requires integrating heterogeneous textual, audio, and visual cues while accounting for conversational context and emotional dynamics. We extend the Self-Distillation Transformer architecture for ERC with appearance+geometry visual representations, class-wise adaptive modality fusion, and a valence-arousal prior for affective transitions. On the MELD and IEMOCAP datasets, geometry-enhanced visual representations improve weighted F1 by 0.27 and 4.36 points over appearance-only features, respectively, while class-wise adaptive fusion provides further gains of 0.17 and 0.25 points over the original softmax gate. The valence-arousal prior yields targeted improvements of 0.30 and 0.74 accuracy points on emotionally shifted utterances while preserving performance on stable turns. These results indicate that structured facial cues, emotion-dependent modality weighting, and affective geometry provide complementary benefits for multimodal ERC.
Read the full abstract
Materials Science
Dimensional Control of Excitonic Interactions in Exfoliated 2D Molecular Crystals
Excited states in molecular crystals depend on intermolecular coupling, charge-transfer interactions, dielectric screening and molecular packing. How reduced dimensionality modifies these interactions when molecular packing structure is preserved remains unclear. Here we demonstrate that tetracene and three other molecular crystals can be mechanically exfoliated into mono-, few- or multilayer flakes, exploiting the top-down approach established for inorganic layered materials. Electron diffraction confirms that exfoliated tetracene retains long-range crystalline order and nearly bulk-like in-plane lattice parameters. This structural fidelity enables systematic thickness-dependent spectroscopy with limited changes in molecular packing. With decreasing thickness, the Davydov splitting diminishes, whereas the Stokes shift increases to nearly three times its bulk-like value, indicating enhanced excited-state relaxation. The photoluminescence 0-0/0-1 intensity ratio also rises, suggesting greater exciton delocalization. Compared with exfoliated tetracene, assembly-grown crystals exhibit lattice contraction and larger Davydov splitting. Analysis within the Frenkel-charge-transfer mixing framework qualitatively accounts for the Davydov-state energy shifts and splitting in terms of molecular packing, intermolecular coupling and dielectric screening. Our results show that intermolecular interactions, excited-state relaxation and exciton coherence can be systematically tuned by varying layer number, demonstrating the feasibility and utility of extending dimensional control of inorganic 2D materials to molecular crystals.
Read the full abstract
Mesoscale and Nanoscale Physics
Apparent Zero-Momentum Signals from Magnon–Magnon Interference in Near-Field Spin-Wave Imaging
Spatially resolved measurements of coherent waves are commonly read as direct maps of the underlying eigenmode spectrum. We show that this interpretation fails in near-field imaging of multimode spin-wave transport. Scanning nitrogen-vacancy magnetometry of yttrium iron garnet reveals pronounced apparent low-wave-vector signals at frequencies for which no propagating spin-wave mode exists. By tuning the probe-sample distance and modeling the anisotropic near-field response, we identify their origin as coherent mixing between directly excited Damon-Eshbach waves and defect- or transducer-scattered spin waves: the measured signal acquires spatial Fourier components at the difference wave vector $\mathbf{k}_{\mathrm{exc}}-\mathbf{k}_{\mathrm{scat}}$, which can vanish even though both constituent waves carry finite momentum. These results establish that near-field spectra of coherent excitations are interferometric field spectra rather than eigenmode maps, providing both a caution and a framework for identifying scattering processes in nanoscale magnonic devices.
Read the full abstract
Analysis of PDEs
Superlinear nonlocal diffusion systems for multispecies populations: well-posedness and discrete chain rules
A nonlocal diffusion system describing the dynamics of multi-species populations and approximating the Shigesada-Kawasaki-Teramoto (SKT) model is analyzed in a bounded domain. The model consists of a system of integro-differential equations of Andreu-Mazón-Rossi-Toledo type, in which a nonlocal cross-diffusion operator acts on nonlinear diffusion potentials. We establish the global existence and uniqueness of strong solutions. The existence proof relies on suitable a priori estimates derived from an entropy inequality, whose proof requires the development of novel discrete chain-rule inequalities. As a by-product, these inequalities provide the basis for the construction of structure-preserving finite-volume schemes for the corresponding local SKT system. Uniqueness of solutions is established by means of a duality argument. Finally, one-dimensional numerical simulations illustrate the nonlocal-to-local limit and the qualitative behavior of solutions for superlinear and sublinear diffusion potentials.
Read the full abstract
Machine Learning
Multi-Pass, Multi-View Blended Learning for High-Fidelity Volumetric CT Synthesis from Chest X-Rays
Reconstructing volumetric Computed Tomography (CT) from a single 2D chest radiograph (CXR) is an ill-posed inverse problem, further complicated by the scarcity of paired CXR-CT training data. Prior approaches address this by training on Digitally Reconstructed Radiographs (DRRs), which are synthetic projections derived from CT volumes. However, the domain gap between DRRs and real CXRs limits generalization, often resulting in coarse or anatomically inconsistent reconstructions when applied to clinical images. To address this challenging problem, this study introduces a Multi-Pass Multi-View Blended Learning framework for synthesizing high-fidelity volumetric CT directly from real chest X-ray (CXR) images. The proposed approach progressively decomposes the synthesis task into two distinct, complementary learning stages. Stage 1 is an unsupervised CXR-to-DRR Domain Adaptation, while Stage 2 includes three passes, namely, (a) supervised DRR-to-CT Transformation, (b) unsupervised Multi-View Slice Refinement, followed by (c) Progressive Transfer Learning (PTL). With such a blended learning paradigm, the proposed approach mitigates the synthetic-to-real domain gap while enhancing both the structural integrity and anatomical detail of the final output. On the LIDC-IDRI dataset, where paired DRR-CT ground truth is available for quantitative evaluation, the proposed method improves upon prior methods by up to 14% in PSNR and 7.6% in SSIM. The framework successfully generates structurally consistent and anatomically realistic high-fidelity CT volumes from real CXRs, marking a significant advancement toward clinical viability of CT reconstruction from standard radiographic images.
Read the full abstract
High Energy Astrophysical Phenomena
Time-Correlated Profile Variability in the MeerKAT Pulsar Timing Array
Long-term pulse-profile variability has emerged as a potentially important source of uncertainty for high-precision pulsar timing, as changes in the integrated pulse shape can bias pulse times of arrival and affect the sensitivity of pulsar timing arrays to nanohertz gravitational waves. We present a systematic search for long-term pulse-profile variability in the six-year MeerKAT Pulsar Timing Array data, analysing observations of 84 millisecond pulsars using a two-dimensional Gaussian-process framework to identify coherent temporal and phase-dependent profile evolution. We detect pulse-profile variability in 18 pulsars. Using the frequency dependence of the observed profile evolution together with scattering simulations, we classify six pulsars as exhibiting chromatic variability consistent with interstellar scattering, while the remaining twelve display predominantly frequency-independent behaviour suggestive of intrinsic changes in the pulsar emission process. Correcting for the S/N-dependent detection bias, we infer that intrinsic profile variability is substantially more common than implied by the detected sample, with at least 40 per cent of millisecond pulsars exhibiting intrinsic profile variability at the 95 per cent credible level. Motivated by the association between pulse-profile variability and spin-down-rate ($\dot{\nu}$) switching in canonical pulsars, we search for corresponding $\dot{\nu}$ variations but find no statistically significant detections. Simulations indicate that the expected variations remain below the sensitivity of current datasets, although unresolved changes may contribute to weak achromatic timing noise. These results suggest that long-term pulse-profile variability may be common among millisecond pulsars and should be explicitly modelled in future high-precision pulsar-timing experiments.
Read the full abstract
Robotics
ViBe: Visual Behavior Adaptation for Perceptive Humanoid Whole-Body Control
Motion tracking provides a scalable recipe for humanoid whole-body control. By design, the resulting trackers lack exteroceptive feedback hence reacting to the environment remains the responsibility of a higher-level planner. Existing perceptive controllers train geometry-only encoders from scratch, trading semantics for sim-to-real ease, and typically rely on teacher-student distillation for a task of interest. We present ViBe, a post-training framework for adapting motion trackers to perceptive control tasks. We leverage pre-trained visual encoders with a multi-query extractor module to learn task-relevant perceptive feedback. This feedback is grafted onto the tracker's input via low-rank adapters, enabling parameter-efficient fine-tuning. Given a task reward and a reference dataset, this modular controller can be adapted directly via policy optimization. Across four tasks, ViBe shows zero-shot sim-to-real transfer spanning perceptive walking on curbs and parkour, Repose Cube, omni-object loco-manipulation, and dodgeball, with visually robust performance across outdoor, low-light, and RGB distractor conditions. Finally, we solve a goal-oriented Repose Cube task with a deliberately simple planner, demonstrating the efficacy of perceptive controllers, adapted by our approach.
Read the full abstract
Number Theory
Unramified Motivic Alternating Multiple Mixed Values
Many variants of the multiple zeta values have been studied in recent years. There are a few among them that remain to be real numbers, such as the multiple mixed values defined by the authors as level-two generalizations, which include both Hoffman's multiple $t$-values and Kaneko–Tsumura's multiple $T$-values. A central question is when such a value descends to level one, that is, when it can be expressed as a $\Q$-linear combination of multiple zeta values. We call these values unramified. In this paper, we further consider the alternating version of the above variants and identify five families of unramified alternating multiple mixed values using the descent theory of Brown and Glanois. We conjecture that all unramified truly alternating multiple mixed values are given in this paper.
Read the full abstract
Applied Physics
Spectral Geometry and Dispersion-Constrained Projection of Diffusive Fields
Diffusive fields obey operator-imposed relations between spatial structure and temporal decay, yet conventional spectral filtering selects components primarily according to frequency or wavenumber magnitude. Here we show that the diffusion operator defines a spectral geometry in the joint space of spatial wavenumber and modal decay rate, where physically admissible modes occupy the manifold $\eta=\alpha|\mathbf{k}|^2$. This geometry separates spectral scale from physical consistency: high-wavenumber modes can remain diffusion-consistent, whereas lower-wavenumber modes can violate the governing dynamics. We exploit this distinction by introducing an operator residual and a finite-width soft projection that selects spectral components according to their distance from the diffusion manifold rather than their spectral magnitude. Numerical studies demonstrate robust recovery under noise, diffusivity mismatch, and finite acquisition, and reveal a consistency–retention tradeoff governed by the manifold width. Photothermal experiments further confirm that the projection suppresses off-manifold spectral content while retaining the dominant thermal response. These results establish operator consistency as a spectral-selection principle for diffusive fields and provide a geometric framework for physics-informed processing of dissipative systems.
Read the full abstract
Functional Analysis
Sharpness Loss for Radon–Nikodym Derivatives of Completely Positive Maps
The Radon–Nikodym theorem for completely positive maps identifies the order interval below a map with positive contractions in the commutant of its minimal Stinespring representation. In this paper, we study the behavior of projection-valued Radon–Nikodym derivatives (sharp submaps) under composition. We prove that sharpness loss is determined by the multiplicative defect of the induced pullback map. In the finite-dimensional case, we show that this defect is determined by the orthogonal complement of the composite Kraus relation space, denoted $\mathcal{E}_{\Lambda,\Omega}$. A sharp submap remains sharp if and only if $\mathcal{E}_{\Lambda,\Omega}$ reduces the lifted Radon–Nikodym projection. Furthermore, we establish a tensor-factor criterion for the global preservation of sharpness and explicitly quantify the sharpness-loss defect.