Papers

Many datasets arise as multivariate time series observed at irregular time intervals, particularly in Earth observation and astronomical measurements. Classical time-series models assume that time is discrete and observations occur at equally spaced intervals, limiting their ability to capture dynamics when observation gaps vary. Existing approaches to irregular sampling rely on continuous-time formulations, which assume that observation intervals are sufficiently small. We address this limitation by introducing a discrete-time framework that accommodates irregular observation gaps while preserving multivariate dependence structures. We introduce a framework for the analysis of irregularly observed multivariate time series based on hypercomplex autoregressive processes. The framework embeds multivariate observations into a hypercomplex algebraic structure, allowing multiple variables and their interactions to be represented within a single mathematical entity while incorporating irregular observation gaps. The resulting process admits a structured matrix representation, which enables the model to be expressed as a state-space system. This formulation provides an estimation methodology in which model parameters can be inferred using Kalman filtering techniques. The state-space representation allows recursive estimation and prediction in the presence of irregular sampling intervals. The performance of the methodology is illustrated through data and applications to remote sensing and astronomical datasets, where observations occur at nonuniform time intervals. The results demonstrate that the approach captures dynamics and cross-variable interactions difficult to model using traditional time-series techniques. The proposed framework provides a methodology for analyzing irregularly sampled multivariate time series and opens possibilities for modeling observational data across disciplines.

Let \(G\) be an infinite residually finite countable discrete group with Kazhdan's property~\((T)\). Assume that its finite-dimensional irreducible unitary representations admit an exhaustive ordering with nondecreasing dimensions tending to infinity and bounded consecutive dimension ratios. From this data we construct a separable unital residually finite-dimensional \(C^*\)-algebra with a faithful quasidiagonal tracial state \(\tau=\frac14\mu_1+\frac34\mu_2\), where \(\mu_1\) is not quasidiagonal. For the resulting algebra, quasidiagonal traces do not form a face of the tracial state space. The construction has two copies of a weighted representation ladder joined at their top levels. We prove that every quasidiagonal trace gives equal weight to the two copies. The proof combines uniform Kazhdan spectral projections, exact ranks of rounded finite-rank compressions, and a weighted Hilbert–Schmidt comparison that controls all representation blocks.

Hardware Construction Languages (HCLs) aim to improve hardware design productivity while generating register-transfer-level (RTL) circuits without changing the designer's microarchitecture. However, most comparisons between HCLs are either qualitative or evaluate quality of results (QoR) across different designs, making it difficult to separate language effects from design effects. This paper compares the most widely used HCLs using the same fixed design, the OCP MXFP4 block dot product, a quantization primitive at the heart of edge Physical-AI inference, implemented as a single 12-stage, II=1 pipeline. A SystemVerilog baseline is followed by implementations in Chisel, SpinalHDL, Amaranth, Clash, Bluespec, and C++ for high-level synthesis (HLS). Every variant goes through the same flow on the same Artix-7 device set at 100 MhZ, driven by a RISC-V soft core. With the micro-architecture held constant, the comparison is clean: every variant meets timing, and the HCLs match or even undercut hand-written RTL in area. The remaining differences stem not from the algorithm but from how each back end lowers arithmetic, and from a single width choice that silently toggles DSP inference. Unlike HLS, where design decisions are limited to pragmas, the HCLs achieve comparable area and timing. Therefore, the choice comes down to ecosystem fit and interface needs rather than QoR.

We construct theories of Carrollian electromagnetism coupled to an axion field. Obtained by performing a $c \to 0$ expansion of axion electrodynamics, these are of two types: electric and magnetic. We derive the Lagrangians and work out the general classical solutions for both. While the electric theory has the axion effectively decoupled from the electromagnetic fields, the magnetic theory is endowed with a nontrivial axion coupling that leads to interesting phenomenology. In particular, a Carrollian analogue of the Witten effect, wherein a magnetic monopole in the presence of a constant non-vanishing axion acquires an electric charge and becomes a dyon, is present in the magnetic theory. At the same time, the linear magnetoelectric effect vanishes entirely, a conclusion we establish by performing a thorough Dirac-Bergmann analysis of the constraint structure of the magnetic theory and a proper identification of its physical phase space.

We investigate a family of $k$-essence cosmological models selected by the requirement that the field equations admit a nontrivial variational symmetry. Inside this general family of theories, we concentrate on the simplest modification to quintessence, induced by power-law terms of the kinetic energy. The symmetry generator introduces a conservation law which we use to obtain an exact vacuum solution. We generalize our analysis and investigate the structure of the phase-space of the cosmological field equations by performing a dynamical systems analysis, both in the absence and the presence of a pressureless matter source. We employ Hubble-normalized variables and the Poincaré compactification, so as to determine the asymptotic behavior of the model and identify all admissible cosmological epochs, from early-time to late-time behavior. We discuss in detail the existence and stability conditions of each fixed point as functions of the free parameters, and characterize the full asymptotic structure of the cosmological history as provided by this symmetry-driven $k$-essence model. We explore the conditions under which the model can describe cosmic acceleration. We find that this $k$-essence model successfully describes the inflationary epoch with a natural exit to the matter-dominated era, with the inflaton ultimately evolving to a dark matter source.

Vision-language navigation requires aligning language with visual observations while maintaining spatial understanding over time. Geometry foundation models (GFMs) expose intermediate representations throughout their hierarchy, but how navigation policies should use these features and retain historical geometric evidence remains unresolved. We introduce \method{}, a streaming VLN framework that addresses these questions across representation depth and navigation time. Hierarchical GFM–VLM fusion couples earlier, intermediate, and later GFM representations to successive policy stages instead of repeatedly injecting a terminal feature. Navigation-aware GFM memory retains historical VGGT global-attention KV states according to instruction relevance, geometric confidence, and transition novelty under a bounded per-layer budget. Retained states provide geometric context for subsequent observations before fusion with the policy. Across R2R-CE and RxR-CE, \method{} achieves strong performance using a single RGB stream without additional navigation-specific external data. Controlled ablations show that multi-depth coupling substantially outperforms repeated terminal-feature injection at matched fusion locations. Bounded navigation-aware retention preserves navigation performance while considerably reducing GFM-KV memory relative to larger-memory temporal retention. These findings support jointly examining the geometric representations exposed to the policy and the historical evidence retained for future inference. Code will be released upon acceptance at https://humanoid-research.github.io/adageovln/.

We present a fixed nondegenerate tetrahedron with a nested largest-dihedral-anglebisection branch that collapses onto a unit edge. The selected dihedral angles to bisect are uniquely largest ones at every bisection step. Some face angles of the branch elements tend to zero, the corresponding diameter-to-inradius ratio diverges, and the diameters of the elements produced do not tend to zero, although every dihedral angle stays uniformly away from both 0 and {\pi} and all face and dihedral angles satisfy a uniform maximum-angle bound. The proof is based on an exact invariant-range argument. This example shows that the largest-dihedral-angle bisection algorithm can produce degenerating tetrahedral partitions.

We introduce here a new approach to the study of isolated chlorophyll pigments with the electronic spectroscopy of their deprotonated anions in a cold environment at 9K. These anions are isoelectronic to the corresponding neutral molecules, therefore their electronic properties are comparable. This comparison is pursued through this article. The electronic spectrum of the Q bands of deprotonated pheophorbide and methyl pheophorbide are recorded by resonant two photon electron detachment with vibrational resolution and show well-resolved bands extending through the visible spectrum in a very similar fashion to their neutral counterparts in solution. The results are compared with data of neutral pheophytin obtained by fluorescence line narrowing spectroscopy at 4K and resonance Raman spectroscopy and found very similar. The theoretical analysis is achieved by density functional (ground state DFT WB97XD, hybrid exchange-correlation functionals and 6-311++G(d,p) basis sets) and time dependent density functional (TDDFT) calculations. The comparison of the active orbitals in the Q transitions shows a great similarity between neutral and anionic forms together with some orbital and related level inversions. The vibronic spectra are simulated by Franck-Condon calculations to assign the dominant transitions of the electronic spectrum. Globally, this is a new perspective in the electronic spectroscopy of chlorophyll pigments that has been opened.

In this paper, we study the effect of an external random field on the localisation / delocalisation of the long-range Discrete Gaussian Chain (DGC), recently studied by Garban arXiv:2312.04536 and Coquille et al. arXiv:2412.15782. We prove that the model is delocalised at every inverse temperature for any polynomial decay $\alpha > 3/2$. This behaviour differs qualitatively from the one of the disorder-free long-range DGC, for which delocalisation at every temperature occurs only for $\alpha > 2$, similarly to what happens for long-range Ising models as already proved by Aizenman and Wehr. Additionally, we provide a more detailed description of the delocalised phase by identifying the growth exponent and scaling limit of the height at the origin for any decay exponent $\alpha > 3/2$.

Consider the random lacunary series $f(z) = \sum_{k=1}^\infty \frac{\xi_k}{\sqrt{k}} \, z^{2^k}$ on the unit disk, where $\{ \xi_k \}$ are independent standard complex Gaussian random variables. We show that almost surely, a.e. $\zeta \in \partial \mathbb{D}$ is a Plessner point of $f$, yet the image of every Stolz angle with vertex at $\zeta$ has asymptotic density zero. This gives a negative answer to questions of Collingwood and Baernstein concerning possible strengthenings of Plessner's theorem. In this example, for a.e. $\zeta \in \partial \mathbb{D}$, the non-tangential range of $f$ at $\zeta$ has zero area. The non-tangential range cannot be much smaller: we show that for an arbitrary holomorphic function on the unit disk, the non-tangential range has Hausdorff dimension 2 at almost every Plessner point.

The entropy rate of a uniformly quantized stationary Gaussian process is the fundamental limit on its lossless compression. It has no closed form, and the classical high-resolution approximation breaks down whenever the spectral density is small on part of the band, as routine smoothing and filtering produce. Here we present two methods for computing the rate. The first is an analytical approximation, obtained from an exact dithering identity and the Kolmogorov-Szegő formula, in which the quantization noise power acts as a floor on the spectral density. The second is a Monte Carlo estimate of the exact rate: a minimum-phase moving-average representation makes the quantized sequence a hidden Markov process whose optimal particle filter is available in closed form, and the filter's log-likelihood on one long sequence gives the rate. The approximation agrees with the estimate to within a few millibits per sample over most of the parameter range for quantization steps up to the signal standard deviation, including processes on which the classical formula fails; for coarser steps the estimate should be used instead. At a step of a quarter of the standard deviation, linear predictive coding comes within 2 to 4 percent of the rate, FLAC within 3 to 32 percent, and general-purpose compressors remain 12 to 89 percent above it.

Deep Learning-based Side-Channel Analysis (DL-SCA) frequently suffers from catastrophic performance degradation across unseen hardware due to printed circuit board routing differences, silicon process variations, and measurement noise shifts. This poster presents the Synthetic Multiple Device Model (s-MDM), a zero-target-trace generative framework designed to improve cross-device portability. s-MDM combines a structured cVAE generator, a Walsh-Hadamard leakage anchor, continuous style modulation, and decoupled leakage-style–domain critics to synthesize virtual source-device profiles offline. Benchmarked on 32-bit side-channel traces (AES_PTv2), s-MDM maps a precise operational boundary: while physical MDM remains superior on identical electrical clones (D4), s-MDM achieves consistently low key rank on the layout/acquisition-shifted Pinata target, where physical baselines are unstable or misaligned.

We establish convergence bounds for deep $V$-learning with horizon $H$. The algorithm fits a scalar value function to targets from executed transitions and selects actions using a predictive model and the value function. For current observed-successor targets with fresh true-kernel outcomes, the conditional mean is $\mathcal{T}^\beta V$, which averages over behavior-policy actions. The Bellman optimality update is $\mathcal{T} V$. We decompose the update error into six residuals: fitting, transition reuse, target construction, replay, action selection, and exploration. Under $L^s$ concentrability, their $L^p$ norms ($p=s/(s-1)$) control expected $L^1$ policy loss. The bound explicitly weights residuals from only the last $H-1$ update blocks, plus an initialization term for shorter runs. We quantify the cost of a shared sampling distribution across horizon levels. For statistical error bounds of order $n^{-\nu}$, we derive optimal continuous allocations and an integer allocation whose objective is within a factor $2^\nu$ of the constrained optimum. A margin condition with exponent $\alpha$ gives action error of order $\Lambda^{1+\alpha/p}$, where $\Lambda$ combines network drift and score error; a one-step construction proves the exponent sharp. Bounds on the distance between frozen and optimal scores transfer an optimal-gap condition to frozen-iterate gap bounds while retaining the mass of optimal ties. Survival probabilities and coverage conditions at deployment yield bounds for policies selected with approximate scores. Separate spatial ReLU networks per horizon level give a conditional neural regression rate, and the finite-state case gives a log-free expected fit rate. These results give expected policy-loss consistency for the fixed-horizon generative-reset approximate-ERM procedure with exact action scores and provide an explicit residual-decay criterion for FIFO/interleaved SGD.

The static screening response of a metallic nanoparticle is encoded in its induced charge density. Integrated observables such as the dipole moment or the polarizability do not determine this density, because many different screening profiles yield the same integrals. We show that in noble-metal nanoparticles the screening charge is atomically structured, with facet, edge, and vertex sites responding differently from a smooth classical conductor. This structure survives in a surface layer of about one atomic width even when the integrated response has reached the conductor limit, and it is absent from continuum descriptions. To resolve it across size, morphology, and composition, we calibrate an atomistic charge-dipole model for Ag and Au nanoparticles directly to first-principles induced-density profiles. The calibration includes the short-range kinetic and exchange-correlation contribution to the hardness kernel that a purely electrostatic model omits, reducing the induced-density-profile error by about a factor of 2.5. A single averaged parameter set per element predicts clusters withheld from the fit and transfers without adjustment to Ag/Au core-shell, alloyed, and elongated particles. A continuous fast multipole implementation of the model scales almost linearly with the number of atoms and reaches several-million-atom particles, far beyond the reach of first-principles methods.

The flip symmetry on link diagrams induces an involution on Khovanov-Rozansky ${\frak{gl}}_{N}$ homology. We prove that this involution is the identity over the integers, generalizing a previous result of Chen and the author. The key ingredients of the proof are (1) a homotopy perturbation argument from a detailed study of the fork twist, which allows us to reduce the computation to planar ${\frak{gl}}_{N}$ webs, and (2) the computation of the flip map for planar ${\frak{gl}}_{N}$ webs via diagrammatics of Soergel bimodules. The latter computation can also be interpreted as a naturality result for the half twist action on type $A$ Soergel bimodules, which might be of independent interest.

Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, limiting their ability to respond to evolving agent capabilities. This disconnect prevents routing strategies from adapting to evolving agent capabilities during post-training and prevents agents from achieving synergistic data-driven specialization. To resolve this, we introduce CERA-MoA (Co-Evolving Router with continually learning Agents for Mixture-of-Agents), an iterative reinforcement learning framework where the dynamic router and independent agent policies co-evolve. We design a predictive familiarity estimator that leverages mid-layer hidden states to evaluate semantic competence among agents, avoiding the overhead of full rollouts. Based on these familiarity scores, a cumulative-threshold adaptive routing mechanism dynamically activates a tailored minimal agent subset, achieving a trade-off between task performance and efficiency. By proactively allocating targeted training samples to agents based on their evolving competence, CERA-MoA promotes capability differentiation. Extensive experiments across various domains demonstrate that CERA-MoA outperforms state-of-the-art static-agent routing and fix-workflow fine-tuning baselines.

BFT accountability is well understood for safety violations, and recent work attributes global liveness violations; recipient-selective silence remains unresolved. A selectively silent adversary withholds messages from some honest nodes while behaving correctly toward others. It can stall consensus yet evade every existing mechanism. We initiate a systematic study of accountability against selective silence. Negatively, a lone attacker silent toward at most $f$ honest nodes is indistinguishable from an honest node, yielding a universal lower bound $K_{\mathrm{SI}} \ge f{+}1$ on the silence identification threshold; moreover, any feedback-free repair after a silence-induced violation costs $\Theta(n^3)$. Positively, Vigil, a Tendermint variant, matches these bounds with attack-adaptive forwarding, via bitmap cross-attestation, core-based membership, and challenge–response auditing. It pays $O(n)$ authenticators per node when no selective silence occurs (plus $\Theta(n^2)$ bitmap metadata bits per node), relays in proportion to the attack's width (sub-threshold silence can force up to $n^3/27$ relays per view, a cost we price exactly), and majority-accuses any node silent toward more than a tunable resilience $\tau_A$ of honest peers ($K_{\mathrm{SI}} = \tau_A{+}1$, optimal at $\tau_A = f$). We also price the residual sub-threshold griefing surface exactly and extend identification to $x$-partial synchrony. Real-network experiments on a three-region WAN, together with a simulator held to exact equality with every closed form, confirm each threshold and cost: at $2\%$ loss, an $f{+}1$ accusation bar falsely accuses $91.2\%$ of honest nodes, while our majority bar accuses $0.002\%$.

In this paper, we study multiple Eisenstein series (MES) in positive characteristic. By computing and analyzing the $t$-expansions of MES, we determine the precise "weights" of their coefficients. This framework enables us to establish a graded algebra structure for MES, extending the direct sum result for Thakur's multiple zeta values proved in [Cha14]. Our result may also be viewed as a function field analogue of the corresponding direct sum result of Bachmann and Kanno [BK26].

Multi-robot navigation in human-shared spaces is inherently interactive: coordinated robot motions influence how nearby entities respond, while those responses provide valuable information for subsequent robot decisions. However, existing methods typically address action-conditioned prediction, multi-robot planning, or online adaptation separately, and therefore lack a unified mechanism for modeling joint robot-entity interactions and adapting future decisions from executed interaction outcomes. To address this gap, we propose TRACER, a bi-directional receding-horizon framework that closes the loop between prediction and adaptation. TRACER evaluates candidate (i.e., alternative feasible future motion plans for the robot team) trajectories using a per-entity probabilistic response model that separates individual-robot effects from non-additive pairwise interactions; after executing the selected trajectory prefix, it updates persistent identity-bound beliefs over latent response modes using the synchronized observed responses. These updated beliefs then guide subsequent candidate evaluation under probabilistic safety and response-aware cost criteria. Experiments show that (i) TRACER more accurately captures non-additive multi-robot interaction effects than a capacity-matched additive predictor, (ii) persistent identity-consistent evidence improves response prediction and downstream replanning, and (iii) the complete TRACER framework improves collision-free completion over an independent-robot baseline on the SocialGym2 multi-robot social-navigation benchmark.

Despite many experimental attempts, laboratory pair-beam sources have not yet reached the charge, collimation, and density needed to observe kinetic pair-plasma instabilities. Here we show that an optimised direct-laser-acceleration (DLA) regime in a gas target can close this gap in the regime of kilojoule-class petawatt lasers. Quasi-3D PIC and Geant4 simulations show that DLA-accelerated electron bunches carry tens of nC at GeV-scale energies with low divergence. Converted in a high-Z target, they generate up to 3x10^12 positrons with tunable energy spectra and transverse sizes of several skin depths in the beam rest frame. A proof-of-principle PIC simulation of the pair beam propagating through background gas shows the growth of a current-filamentation instability. This provides direct numerical evidence that kinetic pair dynamics are within reach of near-term facilities such as ELI-L4. The same electron driver also generates 6.5x10^5 Bethe-Heitler muons per shot, at up to 580 muons/J. We derive a scaling law for the muon yield and validate it against Geant4 simulation. Above a modest driver-energy threshold, the yield depends on the total energy delivered to the electron beam rather than on its peak energy - a design principle for future laser-based muon sources. These results establish DLA-driven secondary sources as a practical, near-term pathway to laboratory pair plasmas, high-yield muon beams, and lepton-accelerator injectors.

We consider a class of neural field models describing the macroscopic activity of an infinite population of neurons organized into finitely many stacked layers, written as a system of semi-linear higher-order parabolic equations. For this class of biologically inspired equations, we show that the higher-order dissipation, modeling gap junction activity regulation, induces a spectral gap ensuring that the associated semigroup possesses an inertial manifold, that is, a finite-dimensional manifold attracting all orbits of the infinite-dimensional system of neural activity. Our main result is an upper bound on the dimension of this manifold, giving its explicit scaling in the dissipation order and strength, the leaking rate, and the norm of the connection operator. We interpret this as a step toward the mathematical modeling of the empirically motivated concept of neural manifold, namely the idea that the collective activity of a large neural population lies on a low-dimensional manifold. For this class of equations, we further prove universal approximation properties, namely stationary pattern expressivity in terms of the input and input-dependent dynamical expressivity over finite time horizons. Finally, we provide sufficient conditions for the existence of traveling waves and nontrivial stationary solutions.

Long-range infrared imaging frequently confronts dense target clusters whose diffraction-limited signatures merge into a single indistinguishable blob, concealing the number, sub-pixel positions, and radiant intensities of the underlying sources. While deep learning has advanced general object detection, resolving such Closely-Spaced Infrared Small Targets (CSIST) remains largely unexplored, owing to a systemic infrastructure void and a fundamental paradigm mismatch. The dominant formulation, which reduces unmixing to a blind, discrete sub-pixel separation, is inherently insufficient: without semantic guidance, the ill-posed inverse problem admits ambiguous solutions plagued by false and missed detections, while grid-based discretization locks predictions onto fixed lattice centers, chaining precision to prohibitively expensive grid refinement. We argue that CSIST unmixing should instead be informed and continuous. To ground this paradigm shift, we establish the first comprehensive open-source ecosystem for the field, comprising the large-scale CSIST-100K benchmark, a tailored metric suite, and the GrokCSO toolkit. Upon this foundation, we propose DISTA-Net++, which anchors a dynamic deep unfolding backbone with two synergistic mechanisms: a Count-Guided Prior that injects the global target count as an explicit semantic constraint to regularize the solution space, and a Continuous Coordinate Rectification that regresses off-grid offsets to decouple localization accuracy from grid resolution. Extensive experiments validate our paradigm: even under the most economical 3x division, DISTA-Net++ surpasses 7x-division state-of-the-art methods by 16.15% in CSO-mAP and 62.96% in count accuracy at merely one-sixth of their computation, demonstrating that unmixing precision need not be purchased with finer discretization. The complete ecosystem is available at https://github.com/GrokCV/GrokDet.

Sign language processing advances rapidly for high-resource languages such as American Sign Language (ASL), yet most of the world's sign languages lack the phonological annotations new methods require. We present the first zero-shot cross-lingual framework for handshape recognition, transferring from ASL to Catalan Sign Language (LSC). Our approach leverages the decomposition of handshapes into five phonological features – selected fingers, flexion, spread, thumb position, and thumb contact – shared across both languages, to decode LSC handshapes from predicted features via a composite phonological distance metric. We evaluate three architectures (MLP, SL-GCN, SHuBERT) trained on two ASL corpora (PopSign, Sem-Lex) against a 37-handshape, single-signer LSC benchmark. Zero-shot transfer proves viable once recording-format disparities are harmonized, reaching 80.0% phonological feature accuracy and 54.5% expected handshape accuracy. Phonological decomposition thus offers a bridge for extending sign language technologies to low-resource languages without any target-language video training labels.

High-entropy topological insulator (Bi3/4Sb1/4)2(Te2/5Se2/5S1/5)3 has been for the first time prepared by self-propagating high-temperature synthesis and melting methods. Single-crystalline thick-filmed sample with thickness of ~0.1 mm was applied to examine the photoresponse properties by using room-temperature chopped-light technique. Photodetector based on (Bi3/4Sb1/4)2(Te2/5Se2/5S1/5)3 demonstrated strong photovoltaic response under irradiation of a white LED light, covering 420-730 nm range. The photocurrent increases from ~20 to ~55 mkA with the increase in intensity from 100 to 600 W/m2 that is originated from the increase in the number of non-equilibrium carriers generated by light. Room-temperature responsivity and detectivity of the photodetector are ~8.7 mA/W and ~109 Jones, respectively.

This paper studies the regularity of the value function arising from a multidimensional continuous-time principal-agent model with separable, nonquadratic effort costs. The associated stochastic control problem has the output and the agent's continuation utility as state variables, and its Hamilton-Jacobi-Bellman equation is fully nonlinear and degenerate, with potentially unbounded coefficients. We address these difficulties by adding an independent regularization noise and bounding the effort. For the resulting problem, we establish classical regularity of the value function and show that the optimal effort is unique, positive and remains in a fixed compact subset, uniformly with respect to both the control restriction and the regularization parameter. These estimates allow us first to remove the control restriction and then to let the additional noise vanish. Consequently, we prove that the regularized value function converges to the original value function and conclude that the latter belongs locally to the Sobolev space $W^{2,1}_{\infty, \mathrm{loc}}$, thereby extending the regularity analysis to separable nonquadratic effort costs, for which the arguments yielding classical solutions in the quadratic-cost setting no longer apply.

Correctly answering a question grounded in normative documents often depends on information outside any single passage: whether the retrieved document is the version currently in force; whether it applies to the jurisdiction, subject (such as an institution or applicant), and date at issue; and whether each normative claim can be traced to its supporting source text. Hosted retrieval services have substantially lowered the engineering cost of building an initial system over such corpora, making "upload the documents and ask" a common default. We evaluate this default on approximately 73,000 candidate normative documents supplied to a production deployment. The evaluation uses a stratified sample of 200 questions from our published benchmark, with a gold source document for every question; the released sampling rule reads no system outputs or scores. We compare the hosted service with a governed system that resolves version and scope through explicit rules before generation. The governed system scored 97.7 overall, while the hosted service scored 88.1, a gap of 9.6 points computed from unrounded means. The question set, the answer text evaluated for both systems, the scores, and the scripts used to reproduce the reported benchmark statistics are public. The governed configuration has operated as a commercial product since January 2026 and serves 1,126 registered users; named customer organizations include Zhipu AI and Lecheng Health. By mid-April 2026, it had reached roughly 100,000 calls per workday.

Collective dynamics in dense cell monolayers are governed by the interplay between crowding and cellular motility. Although increasing density can slow cellular motion and promote glass-like behaviour, the dynamical state of dense HeLa monolayers remains unclear. Here, we combine in vitro time-lapse imaging of HeLa cell monolayers with simulations of a deformable active-cell model to examine how cell density and motility regulate collective relaxation. Within the experimentally accessible density and time ranges, untreated HeLa monolayers remain liquid-like: structural relaxation progressively slows down with increasing density but remains observable throughout the investigated range. Under low-nutrient conditions, cell motility is strongly reduced, and structural relaxation becomes substantially slower. To elucidate the mechanisms underlying these experimental observations, we further performed simulations using a deformable-cell model. The model qualitatively reproduces the density-dependent increase in structural relaxation time and further shows that reducing self-propulsion promotes long-lived caging dynamics at high packing fractions. These results show that dense HeLa monolayers can sustain slow, heterogeneous, yet relaxing collective dynamics under untreated conditions, and indicate that persistent cellular motility is an important factor in maintaining structural relaxation at high density, which may provide insight into the metastatic potential of cancer cells.

We demonstrate dual-modal distributed sensing on an operational 110-kV OPGW during crane maintenance. DAS resolves meter-scale lifting events and matches impulsive events to phone audio, while DTSS quantifies post-reclamping residual strain up to $\sim$398 $\mu\epsilon$, enabling maintenance verification and asset monitoring in transmission grids.

Large audio-language models have shown promise for anti-fraud detection by directly processing speech and reasoning over fraud-related evidence. Their deployment, however, requires predictions to follow a predefined label space and a structured decision protocol consisting of service-scenario identification, fraud detection, and conditional fraud-type classification. Existing fine-tuning and prompt-based approaches typically encode task knowledge, constraints, and decision rules into model parameters or manually maintained prompts, making them difficult to adapt as fraud patterns and labeling policies evolve. To this end, we propose FRAUDSkill, a structured frozen-weight adaptation framework that leaves the underlying audio-language model unchanged while optimizing an external layer of skill programs, route-specific policies, and decision rules. We further combine structured output control with validation-guided multi-path inference to ensure protocol-compliant predictions. On the TeleAntiFraud benchmark, FRAUDSkill achieves 73.50% Macro-F1, outperforming the shared frozen-model baseline by 31.96% while reducing invalid outputs to 1.94%. Extensive experiments demonstrate that external skill optimization provides an effective and adaptable solution for structured audio anti-fraud detection without modifying the underlying model. The source code is available at https://anonymous.4open.science/r/FRAUDSKILL-114514.

Long-range (LR) interactions are no longer just a theoretical idea: they can be engineered in several low-dimensional platforms and offer a new way to control heat transport. At the same time, they push beyond the standard picture developed mainly for short-range, momentum-conserving lattices. In this review, we consider one-dimensional Fermi-Pasta-Ulam-Tsingou (FPUT)-type lattices where LR effects enter mainly through a quartic anharmonic coupling that decays as a power law, characterized by an exponent {\sigma}. Our goal is to explain, in a unified way, how LR anharmonicity changes microscopic energy exchange, collective dynamics, and the macroscopic scaling laws of heat conduction-while also clarifying what is well established and what is still debated (For more details of the abstract, please see the main text and paper).