Papers

Luminous red novae (LRNe) are potential signatures of common envelope evolution (CEE), a critical yet poorly understood phase in compact binary formation. AT 2025abao, an LRN in the Andromeda galaxy, originated from an asymptotic giant branch (AGB) progenitor and exhibited a prolonged pre-outburst transient. Lacking direct constraints on the companion, however, the system's evolutionary fate remains unclear. Here, we unravel the binary nature of AT 2025abao by modeling its complete evolution, from precursor to final outcome. We show that the eight-year infrared precursor arises from a $\approx1.4^{+0.2}_{-0.3}\,M_{\odot}$ main-sequence companion inspiralling through the wind of a $\approx 6.9\,M_{\odot}$ AGB star. As CEE commences, the companion plunges into the envelope and launches a shock. Radiation hydrodynamic modeling of this outburst constrains the total ejecta mass to $\approx 0.92\,M_{\odot}$ and energy to $\approx 1.25\times 10^{47}\,$erg, revealing that only $\approx 0.57\,M_{\odot}$ becomes unbound. Subsequent radiative transfer simulations successfully reproduce the evolving spectral energy distribution. Crucially, our models indicate the main-sequence companion and AGB core survive in an eccentric orbit without merging at least 400 days after CEE onset. AT 2025abao may therefore be the first observed CEE event caught forming a close compact binary, bridging a long-standing observational gap in compact binary formation.

We ask: given a retrieved source audio $S$ and a separate reference audio $R$, can we synthesize novel audio $Y$ out of this pair $(S,R)$ such that $Y$ remains acoustically consistent with $S$, while not persistently copying segments of $S$ or $R$? The first clause is a well-known goal in Foley audio production, and the second is a well-known issue in neural RAG when $S$ and $R$ are naively injected into neural generators. We show that both clauses can be addressed simultaneously using a method we coin relational synthesis, a variation of concatenative synthesis where target cost is replaced by a relational Gromov-like structural cost. Rather than imitating the content of $R$, relational synthesis exploits it from the "other side of the hill": it transfers the temporal structure and directed amplitude motion of $R$ to reorganize and concatenate the grains of $S$ in a novel manner that protects $S$'s acoustic information. Our experiments show that relational synthesis integrates naturally with neural RAG and produces Foley audio that performs well on metrics measuring temporal agreement, acoustic fidelity, and leakage persistence, while maintaining distribution-level quality and text alignment.

We construct computable error estimators, which double as loss functions for neural networks, for a class of parametric nonlinear partial differential equations with a monotonicity property, and prove that they are globally reliable and efficient. The value of such a loss function is bounded above and below by the squared error in the natural trial norm, for every trial function, not merely for those near the exact solution; this global property rests on monotonicity. The construction rests on splitting the nonlinear operator into a linear part and a strongly monotone closure, and on measuring the linear residual in a discrete dual norm. Since the closure contributes a dual norm that admits no closed form when the trial norm is stronger than an $L_2$ norm, the estimator is built around a computable surrogate for it, required only to satisfy a pairing bound and a Lipschitz bound. The main theorem then yields two-sided bounds with explicit constants and covers two instances. A first-order system least-squares estimator on conforming trial spaces is the first instance studied: its two-sided bound holds on the whole trial space and therefore applies to arbitrary approximations, including those not in a discrete finite element space. The second instance is a discontinuous Petrov-Galerkin estimator on trial spaces of finite element functions for each parameter value, built with broken test spaces, for which no conformity is required and the dual norm is computed by independent element-local problems. All assumptions are verified for a model class of nonlinear fluxes, with the constants tracked explicitly in terms of the parameter range. Being computable for an arbitrary input, both estimators serve as variationally correct loss functions for neural network approximations of parameter-to-solution maps.

For quantum random number generators (QRNGs), strong security guarantees can be obtained under various assumptions. The most well-known assumptions are the non-signalling assumption (needed for device-independent (DI) QRNG) and post-quantum computational assumptions (needed for computationally bounded single-prover QRNG protocols). In this work, we study a third model: a simple prepare-and-measure QRNG whose security relies on physically motivated energy constraints. This physically motivated set-up allows for easier implementations of QRNGs and can supply better rates. So far, previous works that studied the energy-constrained model either assumed that the adversary is classical in their security proof or supplied relatively weak bounds on the certified randomness, by bounding the conditional min-entropy. In this work, we prove a new, essentially tight, lower bound on the single-round von Neumann entropy against adversaries holding both classical and quantum side-information. Our work reveals that working with the von Neumann entropy, rather than the min-entropy, in the quantum case can lead to a major improvement, nearly matching the classical bound in many cases. We further study the robustness of our entropy bound. Since the single-round von Neumann entropy is the main quantity of interest when characterizing the key rate of QRNGs, our bound establishes the main ingredient for rigorous security proofs against general quantum attacks in this framework.

Goal-conditioned imitation learning (GCIL) with flow matching is a promising framework that can represent multimodal behaviors while adapting to diverse, user-specified goals, yet often fails when goals lie outside the demonstration support. To extrapolate to such unseen goals without collapsing multimodality - a problem we call distributional extrapolation - we introduce Bilinear Flow Policy (BFP), a generative visuomotor policy that combines transductive retrieval with a bilinear conditional flow. Given an unseen observation-goal pair, BFP retrieves an "anchor" training example and transductively reformulates the unseen pair as this familiar anchor plus a residual term. For this decomposition to guide action prediction, the residual must compactly encode how the current observation-goal pair differs from the anchor, and the anchor must be chosen so that this difference is predictive of the corresponding action distribution. BFP achieves this with pretrained visual features and a novel learned anchor-selection algorithm. The novel bilinear flow then models how the anchor and the residual jointly determine the multimodal action distribution. We prove that, for bilinear flow under suitable assumptions, action distribution error at unseen goals is bounded by the in-distribution flow-matching error up to problem-dependent factors. Across five manipulation tasks in simulation, BFP achieves 2.63x the out-of distribution success rate of a GCIL policy and 1.36x that of the strongest extrapolation-targeted baseline. On two real-world tasks, BFP improves over GCIL by 32%. Finally, our theory yields practical, pre deployment diagnostics for predicting which trained policies will extrapolate well and to which unseen goal.

Semantic communication has recently gained traction for its ability to reduce the amount of data transmitted over a communication link by transmitting a task-oriented representation instead of the raw source. Zero-shot semantic communication sends a general embedding from a vision-language model (VLM), so the same transmitter can serve new classification tasks without retraining. Most evidence for this advantage, however, comes from numerical simulation. We implement zero-shot semantic communication on a software-defined radio platform: a Raspberry Pi drives a pair of Analog Devices Active Learning Module (ADALM)-Pluto transceivers, with an image encoder at the transmitter and a text encoder at the receiver, and determines the zero-shot classification results via cosine similarity. We compare two VLMs, CLIP and MobileCLIP, across various channel conditions, i.e., different signal-to-noise ratios (SNRs). We validate that the semantic link spends 9x fewer channel uses per image than a JPEG plus 16-ary quadrature amplitude modulation baseline and still reaches 82% accuracy on CIFAR-10 at 22.3 dB, where the baseline scores 0%. On the traffic sign recognition dataset (TSRD), MobileCLIP correctly classifies 98.3% of unseen images at the same SNR. Offloading the image encoder to a neural processing unit reduces encoding to 13.4 ms per image, 49x faster than a Raspberry Pi 4 CPU, placing the transmitter within a real-time budget. Our implementation is publicly available at https://github.com/thanhlexyz/zsscsdr.

We show that the expressions for the matter Lagrangian \(L_m\) and the metric variation \(\delta T_{\mu\nu}\) of a perfect fluid obtained in previous studies appear to be inconsistent with the standard energy-momentum tensor under general conditions. Consequently, a large number of studies in astrophysics and cosmology relying on these expressions may need to be re-examined. By performing a series of straightforward calculations directly on the standard energy-momentum tensor \(T_{\mu\nu}\) together with the particle number conservation condition, we derive an expression for \(\delta T_{\mu\nu}\) that is independent of the choice of \(L_m\). Applying this result to \(f(R,T)\) gravity, we obtain the exact form of the tensor \(\Theta_{\mu\nu} = g^{\sigma\rho} \frac{\delta T_{\sigma\rho}}{\delta g^{\mu\nu}}\), which remains an important yet long-standing controversial quantity. This expression is shown to hold also for radiation, regardless of whether particle number is conserved. A major result is that if the energy-momentum tensor \(T_{\mu\nu}\) of the Universe consists solely of standard components (baryonic/cold dark matter, radiation, and the cosmological constant), then \(f(R,T)\) gravity satisfies the conservation law \(\nabla_\mu T^{\mu\nu} = 0\) for any function \(f(R,T)\). This contrasts with previous studies, which found that the conservation law holds only for a restricted class of \(f(R,T)\) functions. Applying the same formalism to stellar interiors, we derive a class of functions that preserve the conservation law. We construct a specific \(f(R,T)\) model that is consistent at both cosmological scales and in high-density objects such as neutron stars. Remarkably, the same parameter set in this model simultaneously alleviates the Hubble tension and reproduces the observed mass-radius (M–R) relation of neutron stars.

Body-body interactions have long been investigated to leverage the performance of various fluid-based machines. This study analyzes the flow interaction between a cylinder and a NACA0012 airfoil using direct numerical simulation and machine learning. We consider a two-dimensional incompressible flow at Reynolds number 100, where the airfoil is positioned in the wake of a cylinder. By sweeping the parameter space, composed of the distance between the cylinder and the airfoil $\Delta x$, the relative height $\Delta y$, and the angle of attack of the airfoil $\alpha$, four characteristic vortex-shedding regimes are observed. This study particularly focuses on suppressed vortex shedding at a certain arrangement as oscillation death in coupled oscillators, with the theoretical prediction of flow stability on a machine-learned low-order coordinate. Independent flows around each body are regarded as a system that exhibits isolated periodic vortex shedding at infinite distance, and interactions appear at smaller distances. We employ a synchronization-theoretic autoencoder to derive models for two coupled oscillators from an aerodynamic coefficient. The proposed approach, which enables the assessment of system stability and the prediction of vortical interaction behavior, may pave the way for the analysis of unsteady body-body interactions from the data-driven and synchronization perspectives.

Transverse optical rectification has been reported in a one-dimensional asymmetric Au grating even though an ideal profile $h(x)$ retains the $xz$ mirror $\Me:y\to-y$. Because $J_y$ is odd under $\Me$, whereas the $p$-polarized bilinears $|E_x|^2$, $|E_z|^2$, and $\mathrm{Re}(E_xE_z^*)$ are even, the transverse dc current requires a distinct mirror-odd factor $\Lamy$. We formulate a symmetry-complete effective model combining the origin-independent cell invariant $\Astr=\mathrm{Im}(h_1^2h_2^*)$, Drude–Edelstein spin polarization, and spin–orbit conversion. We then propose a dynamical realization of $\Lamy$ for a finite-width device. A driven pump-like plasmonic mode at $k_y\simeq0$ generates a mirror pair of transverse surface-plasmon-polariton quasinormal modes $(K_x,\pm q_y)$ through the same third-order susceptibility that produces self- and cross-Kerr shifts. Four-wave mixing creates the pair and locks its phase sum but, by itself, cannot change the pair-population difference. Weak coherent backscattering makes the relative drive phase observable; a small phase difference from incident $k_y$, spot displacement, edges, or disorder then unfolds a Kerr directional instability and deterministically selects its sign. The resulting population imbalance supplies the dynamical mirror-odd factor $\Ldyn=\lambda_\eta\eta$ in the structure-dependent $\Astr\Lamy\langle j_xs_y\rangle_t$ side-jump-like channel, which vanishes for the centered symmetric grating because $\Astr=0$. We give the complete three-mode equations, the pump-only and saturated-state Bogoliubov problems, the imperfect-pitchfork reduction, and a validated normalized numerical realization. The normalized equations exhibit a robust supercritical pitchfork with an approximately one-half critical exponent and phase-reversed branch selection.

We construct the operator product expansions (OPEs) of the ${\cal N}=8$ ${\rm w}_{1+\infty}$ currents (obtained from the antiholomorphic light transform) with the matter fields in the ${\cal N}=8$ supergravity. The corresponding (anti)commutators are obtained from the double contour integrals using the above OPEs. By computing the triple (anti)commutators consisting of current-current-matter (i.e., the Jacobi identity for fixed last element), the previously known ${\cal N}=8$ celestial soft symmetry algebra between these currents (in the antiholomorphic sector) is reproduced. We repeat the above procedure by taking the holomorphic light transform and the OPEs of the ${\cal N}=8$ ${\rm \bar{w}}_{1+\infty}$ currents with the matter fields and the corresponding (anti)commutators are obtained. We also obtain another type of ${\cal N}=8$ celestial soft symmetry algebra between these currents (in the holomorphic sector). As a subalgebra in the antiholomorphic and holomorphic sectors, there exists the ${\cal N}=8$ extended BMS (Bondi, van der Burg, Metzner, and Sachs in 1962) algebra. We analyze the above results from the ${\cal N}=8$ conformally soft ${\rm w}_{1+\infty}$ (or ${\rm \bar{w}}_{1+\infty}$) currents in the antiholomorphic (or holomorphic) sector. The nontrivial soft theorems for the graviton, the gravitinos and the graviphotons are described. Finally, using the shadow transform (the consecutive composition of holomorphic and antiholomorphic light transforms), we obtain the OPEs of the ${\cal N}=8$ shadow ${\rm w}_{1+\infty}$ (or ${\rm \bar{w}}_{1+\infty}$) currents with the matter fields where the stress-energy tensor, the supercurrents and the supertranslation generator of the ${\cal N}=8$ extended BMS algebra are identified.

In this paper, we introduce a class of function-preserving neural network operators designed to preserve a prescribed strictly positive function. We examine the approximation behaviour of the proposed operators on the space of continuous functions and show that they are positive, linear, well-defined, and bounded. Specifically, we demonstrate uniform convergence on compact intervals and get quantitative estimates of the approximation error in terms of the modulus of continuity. To support the theoretical analysis, we present several numerical experiments focusing on the preservation of exponential and logarithmic functions. These examples show how well the proposed operators perform and compare them with existing neural network operators. Also, an application to denoising modulated signals is provided, demonstrating the practical relevance of the proposed approach.

Holes in LiDAR scans of environments with glass windows remain an unresolved problem in three-dimensional reconstruction. This study presents a pipeline for point cloud acquisition, filtering, completion, and surface reconstruction to address sparse sampling and missing window regions in scans of a high-speed train nose. FAST-LIVO2 provides the initial point cloud through multisensor odometry and mapping, and moving least squares (MLS) smooths the observations. We then introduce three-axis projection-based subdivision and interpolation with reverse hole boundary identification, referred to as three-axis reverse completion. The method interpolates missing regions from observations around each hole. Greedy projection triangulation, Poisson surface reconstruction, and a Marching Cubes-based pipeline generate meshes from the completed point cloud. Experiments on a proportionally scaled display model of a high-speed train nose show that the proposed method fills missing point cloud regions around the glass windows. Under the evaluation setting used in this study, greedy projection triangulation yields lower geometric distance errors than the other two reconstruction pipelines. The pipeline supports non-contact digital modeling of train nose geometry and provides a practical approach to reconstructing objects with glass windows.

This paper introduces Hybrid DeepSEE (HDS), a Human-in-the-Loop (HITL) neuro-symbolic framework for proactive drift anticipation in Visual SLAM (V-SLAM). While data-driven models offer predictive power, their "black-box" nature often yields physically inconsistent outputs in out-of-distribution (OOD) environments. To address this, HDS integrates neural drift risk estimation with symbolic constraint reasoning. By utilizing a Large Language Model (LLM) as a reasoning bridge, the framework translates qualitative human context into interpretable symbolic constraints. Building upon this architecture, we propose a superior drift anticipation framework that ensures enhanced reliability and consistency in Visual SLAM

In the rapidly evolving field of construction management, worker safety remains a top priority. This paper introduces an innovative vision-based system for real-time detection of helmet compliance, specifically designed for construction sites, utilizing advanced computer vision techniques and machine learning algorithms within the YOLO (you only look once) framework. Our system leverages high-resolution video feeds from strategically positioned cameras to monitor adherence to safety regulations regarding helmet usage. By employing deep learning methodologies, the system effectively identifies individuals not wearing helmets, thereby significantly mitigating the risk of head injuries among workers. Our training and validation results revealed an impressive precision exceeding 97% at mAP@0.5 for both helmeted and non-helmeted individuals. Furthermore, our experiments demonstrate exceptional detection accuracy, demonstrating the system's resilience under varying lighting conditions and diverse worker movements. The consistent decrease in loss and improvement in metrics throughout training validates the effectiveness of the YOLOv8 model in enhancing recognition performance. The implications of this research extend beyond mere regulatory compliance, opening avenues for innovative applications in occupational safety management. This study highlights the critical role of technology in protecting lives and lays the groundwork for future advancements in smart construction environments.

Diffusion and VLA policies for manipulation are often deployed through downstream impedance controllers. The stiffness and damping gains of these controllers affect task success, yet are commonly inherited from data collection rather than selected for the deployed policy. Although empirical gain sweeps can improve performance, they require repeated evaluation rollouts. We introduce PHRetune, an offline method that derives controller gains for a frozen policy without evaluation rollouts or gain search. Our approach learns a port-Hamiltonian model from demonstrations to estimate the effort and energy associated with the policy's predicted actions. The policy is applied to recorded demonstration observations, and its predictions are assessed against demonstration-derived effort and energy budgets. From this comparison, we derive a single gain scale in closed form, adjusting the downstream controller while preserving the policy and its action representation. The gains are fixed before evaluation, without requiring a prior manipulator model, task rewards, policy retraining, or additional runtime computation. Across LIBERO suites, PHRetune improves Diffusion Policy success by up to 9.4 percentage points, with the derived gains achieving the highest observed success rates in empirical gain sweeps. On all four real-world manipulation tasks, PHRetuned Diffusion Policy outperforms the nominal policy, alternative gain-tuning methods, and a policy-retraining baseline. The same procedure improves success with SmolVLA and OpenVLA-OFT on every task, while reducing acceleration and jerk for both VLA backbones.

We study a dynamical optimal transport problem on path-space with kinetic cost and nonlocal interaction. For a Gaussian interaction kernel, cyclical monotonicity and the regularity of the effective endpoint cost yield a conservative vector field associated with the initial momentum of the optimal trajectories. Separately, the Gaussian interaction potential $W_{\pi_0}$ satisfies a uniform gradient estimate and, under a quantitative condition on the interaction strength, is a classical and hence viscosity subsolution of an eikonal equation. We also derive the Euler–Lagrange dynamics directly from path-space optimality by means of endpoint-preserving perturbations of the optimal path measure. The associated phase-space marginals satisfy a measure-valued Liouville-type, or nonlocal kinetic, equation driven by the smooth Gaussian self-consistent force. Conversely, a superposition principle lifts suitable measure-valued solutions of the phase-space continuity equation to measures on phase-space trajectories. Thus path-space optimal transport provides a variational connection between endpoint geometry, eikonal behavior of the Gaussian interaction potential, and kinetic mean-field dynamics.

Let $n\geq2$, $s\in(0,1)$, and let $\Omega\subset\mathbb{R}^n$ be a bounded Lipschitz domain. In this paper, we establish optimal global Sobolev estimates for the fractional Dirichlet problem \begin{equation*} \left\{\begin{aligned} (-\Delta)^su&=f & & \text{in}\ \ \Omega, u&=0 & & \text{in}\ \ \mathbb{R}^n\setminus\Omega. \end{aligned}\right. \end{equation*} More precisely, for any given $t\in[s,\min\{2s,\,s+1/2\})$, we prove that there exists a small positive constant $\eta=\eta(n,s,t,\Omega)$ such that the following holds. If $n=2$ and $s\in(1/2,1)$, then, for any $q\in(1,\frac{4}{1+2s}+\eta)$ and $p\in(1,\frac{nq}{n-(2s-t)q}]$, \begin{equation*} \|u\|_{W^{t,p}(\mathbb{R}^n)}\le C\|f\|_{L^q(\Omega)}, \end{equation*} and, if either $n=2$ and $s\in(0,1/2]$ or $n\ge 3$ and $s\in(0,1)$, then, for any $q\in(1,\frac{n(2s+1)}{nt+(2s-t)(2s+1)}+\eta)$ and $p\in(1,\frac{nq}{n-(2s-t)q}]$, \begin{equation*} \|u\|_{W^{t,p}(\mathbb{R}^n)}\le C\|f\|_{L^q(\Omega)}. \end{equation*} Here the positive constant $C$ depends only on $n$, $s$, $t$, $p$, $q$, and $\Omega$. The stated universal baselines are optimal in Dahlberg's sense: for any larger target exponent $p$ there is a bounded Lipschitz domain whose solution with right-hand side identically $1$ is not in $W^{t,p}(\mathbb{R}^n)$. The counterexamples are constructed using regular Cantor sets on the boundary and, in the planar exceptional case, homogeneous solutions in sectors.

Consider a quantum state vector proportional to any given column of the character table of the symmetric group ($S_n$), corresponding to a superposition over the irreps weighted by the character values. It was recently argued that sampling from this state is classically hard under reasonable complexity-theoretic assumptions, while there exists an efficient quantum algorithm for preparing this character state using the quantum Fourier transform (QFT) over $S_n$ [arXiv:2501.12579]. We give an alternative, simpler algorithm to prepare this character state with a sequence of Hamiltonian evolutions, as well as an efficient implementation with reconfigurable qubits that uses only nearest-neighbor gates in 1D. The gate complexity, accounting for Hamiltonian simulation errors, depends on the choice of the column: For fixed target state error, the estimated sufficient total gate count ranges from as few as $O(n)$ gates for the $n$-cycle column, to $\widetilde O(n^{1.75})$ gates near the conjecturally classically hard parameter regime, to provably $\widetilde O(n^{2.5})$ gates in the worst case. In contrast, the QFT subroutine in the prior approach uses $\widetilde O(n^3)$ gates. Our gate count estimates come from leading-order state-dependent Trotter error analysis, and the proven worst-case bound relies on a QSVT-based Hamiltonian simulation method; numerical benchmarks of the Trotter-based protocol suggest that the estimates are conservative. Furthermore, we generalize our protocol to give an explicit algorithm for the quantum character transform (QCT) over $S_n$ using $\widetilde O(n^{2.5})$ gates and $O(n)$ qubits. An application to entanglement entropy of symmetric orbifold conformal field theories is also discussed.

Plane-wave ultrasound enables ultrafast imaging but remains constrained by a receive-aperture-dependent resolution- contrast trade-off. Fine structures benefit from high spatial resolution, whereas diffuse tissue and lesions often require improved contrast and clutter suppression. These objectives favor different receive-aperture configurations, yet conventional beamformers operate at a fixed imaging point. We propose a physics-grounded, objective-conditioned deep beamformer that enables controllable image formation within a single network. From single-plane-wave channel data, the network predicts spatially varying receive-apodization weights, conditioned on the desired imaging objective through feature-wise linear modulation. The conditioning variable is explicitly tied to a physically defined receive-aperture configuration and supervised using the correspondingcoherent-plane-wave-compounding target. Differentiable, region-of-interest-aware, physics-guided quality losses further promote objective-specific resolution and contrast during end-to-end training. The model is trained exclusively on simulated channel data spanning diverse target morphologies and echogenicities, and evaluated zero-shot on the PICMUS and CUBDL datasets, including unseen probe geometries. The proposed framework generally improves upon single-plane-wave delay-and-sum reconstruction while providing objective-dependent control. On PICMUS, the proposed method improves gCNR by up to 16.8% and axial resolution by 22% compared with the reference beamforming method. Moreover, switching between the contrast- and resolution-oriented objectives enables controllable changes of up to 4.9% in lesion contrast and 11.4% in lateral resolution. Compared with the best-performing learning-based baselines, the proposed method reduces lateral full-width-at-half-maximum by 5.9% and improves generalized contrast-to-noise ratio by 6.6%.

Modern information systems, including many agentic workflows, use dense retrieval to explore large amounts of unstructured data. However, dense retrieval relies on surface-level semantic similarity, which is insufficient for increasingly complex search applications. Here, we investigate agentic retrieval that combines the reasoning capabilities of Large Language Models (LLMs) with the efficient corpus exploration of retrievers in a ReAct agentic loop to solve complex retrieval tasks. In our experiments, we show that agentic retrieval is more effective than standard retrieval, improving nDCG@10 by 8.7 points using the same embedding model. Moreover, while specialized retrieval methods struggle on out-of-domain tasks, agentic retrieval is highly generalizable: the same pipeline achieves competitive results on both the ViDoRe v3 and BRIGHT leaderboards. However, this improvement comes at a cost. On average, agentic retrieval takes 107.4 seconds, compared to 0.67 seconds for standard retrieval, and consumes 764.1K input and 5.8K output tokens per query. In short, our study demonstrates the effectiveness of agentic retrieval in modern data systems and motivates future work on more cost-efficient retrieval agents for large-scale deployment.

This article introduces an innovative system that integrates landmark identification with deep learning to enhance fall detection accuracy and reliability. By utilizing advanced computer vision techniques, such as Media Pipe for spatial recognition, the system effectively differentiates between routine movements and actual falls. The integration of landmarks with a deep learning prediction algorithm minimizes false alarms, ensuring timely responses to genuine falls. Comprehensive experimentation underscores the system's versatility across various scenarios, emphasizing its potential to improve safety and independence for older adults. The training process demonstrates a steady increase in accuracy, stabilizing by the 40th cycle, while error rates decline significantly during the initial cycles. Real-time experiments, involving both male and female participants aged 8 to 50, recorded a remarkable 95% detection rate of falls, demonstrating the system's effectiveness and promising future applications in elder care and smart health monitoring environments.

Large language model (LLM) serving scales its replica count with the request load, yet GPU memory still stands idle inside the replicas. Adding a replica takes minutes, while the memory that a replica needs changes within seconds. Even instant autoscaling could not return this idle memory, because the smallest unit that it can remove is a whole replica. Colocating parameter-efficient fine-tuning (PEFT) with inference can use this memory, but inference must be able to reclaim it within seconds, before requests that wait for memory exceed their latency service-level objective (SLO). Existing colocation systems either keep the tuning memory resident or let inference reclaim it at the coarse granularity of a whole training sample. Each such reclamation also discards the running tuning step. To address these limitations, we present MOLT, a fine-grained memory sharing system that lets inference reclaim the memory of individual activations that a running tuning step has saved for its backward pass. The step continues, and its backward pass recomputes those activations. Inference reclaims only memory that no in-flight GPU work can still access, even under CPU–GPU asynchrony and tensor parallelism. On four model deployments (24B–70B) across H100 SXM and B200 GPUs under trace-driven workloads, MOLT keeps inference SLO attainment at or above 99.7% and completes 1.9–3.3x the tuning work of discard-based memory sharing.

Training-free acceleration of a diffusion policy is accepted when an internal similarity signal reports that the shortcut changed nothing. We price every monitor a control loop can afford in closed-loop success rather than feature distance, over $114$ accelerator configurations and four policy families: two accelerators each clear their own gate's bar and log every reuse as certified, while on one task one finishes every episode and the other none. A monitor is two designs, not one — the statistic it reads, and what it does when that statistic fires. Published gates re-arm after every rejection, and under that response even an oracle handed every forward pass and the exact local action error loses twenty points; absorbing the same statistic costs one point, and most of its speed. What makes a response that never forgets affordable is a statistic that rarely fires, and what it must measure is deviation from the policy the accelerator replaced — which a chunked policy has already paid for, its last plan not yet expired and free to read. Guarding every call this way cuts per-call compute by $1.55$–$3.09\times$, where any reference-requiring check at the same coverage would have to stop accelerating altogether. On three of our four families the schedule alone already holds the pre-stated $\pm2$-point margin. What the monitor is measurably worth shows in three places: on the fourth family, where it rescues the candidate selection landed on; on four configurations it did not select; and on a contact-rich fifth family, chosen where the schedule was expected to fail and run after every design choice was frozen, where no unmonitored arm at its speed holds the margin and the monitored one does.

We propose a learning-to-optimize framework for solving parametric mixed-integer nonlinear programs arising in model predictive control and trajectory optimization. The proposed method combines neural network prediction with a sequential quadratic programming (SQP) refinement layer. First, neural networks are trained using a combination of supervised and self-supervised losses to predict integer solutions together with compatible continuous warm starts. To account for prediction inaccuracies, the trained predictor generates multiple candidate integer solutions, each paired with a continuous initialization. The integer solutions are then fixed, and a differentiable SQP layer refines the associated continuous variables over a prescribed number of iterations. The SQP-layer parameters, including the curvature regularization and step size, are trained to improve both feasibility and objective value after every iteration. By separating integer prediction from continuous optimization-based refinement, the proposed framework improves robustness to imperfect neural predictions while maintaining a predictable computational budget. We demonstrate the proposed framework on contact and trajectory optimization for a planar box-pushing example.

Post-training quantization has been shown to preserve VLA performance under standard evaluation conditions, but whether it preserves the full-precision model's robustness and generalization remains underexplored. In this study, we systematically study the robustness and generalization of post-quantized VLA policies under environmental disturbances. Empirical results show that quantized policies can become fragile to subtle environmental variations despite retaining comparable in-distribution performance. We further observe that action discrepancies are concentrated in a small subset of rollout states, while teacher guidance has opposite effects depending on discrepancy: it improves generalization at high-discrepancy states but can degrade it at low-discrepancy states. These findings reveal that effective post-quantization recovery requires selectively intervening on vulnerable states rather than globally distilling the student. We therefore propose Policy-Induced Vulnerability-Oriented Tuning (PIVOT-Q), a vulnerability-aware On-Policy Distillation (OPD) framework that selectively corrects vulnerable states encountered during quantized-student rollouts using the frozen full-precision policy as a teacher. PIVOT-Q identifies vulnerable states using discounted accumulated discrepancies over a short horizon, applies phase-balanced sparse supervision, and uses a Behavioral Anchor to prevent unnecessary changes. Experiments under seven LIBERO-Plus environmental variations demonstrate consistent recovery across multiple VLA backbones and quantization methods. Notably, PIVOT-Q consistently outperforms full-state distillation across all settings while using only 7.4% of its state-level distillation budget. Our code is available at https://github.com/ruanruan-andy/PIVOT-Q.

Mixture-of-Experts (MoE) has been widely adopted in recent large language model (LLM) architectures. However, scaling up MoE in LLM training introduces system-level challenges on training, where non-uniform token routing can lead to highly imbalanced workloads across experts and devices, further destabilizing the training process. With trillion-scale LLMs, imbalanced expert workloads further amplify the resource cost of MoE training, resulting in degraded training efficiency and hardware utilization for underloaded experts, while hot experts require additional resources to accommodate excessive workloads. Recent studies address imbalanced MoE training through intricate parallelism strategies or resource reallocation. However, these system-level approaches often introduce additional resource requirements and considerable orchestration complexity, which become increasingly difficult to afford when training trillion-parameter LLMs under constrained computational resources. This work introduces CIPHER-MoE, which mitigates MoE workload imbalance while keeping the router's token-side Top-K selection unchanged. CIPHER-MoE applies affinity-aware Expert-to-Token filtering with explicit capacity control to reduce hotspot expert workloads without additional hardware resources or complex runtime design. The proposed method has been evaluated on large-scale MoE models, including DeepSeek-V4-Pro, showing up to 64.9 percentage points Top-1 expert workload reduction and 1.10$\times$-1.94$\times$ training acceleration, while preserving the training quality. The source code will be released soon.

RANSAC scoring and local optimization (LO) impose different robustness requirements, motivating the separation of hypothesis selection from refinement. We systematically isolate the effects of robust-loss shape, incorrectly specified inlier scales, and optimization strategy on essential matrix, fundamental matrix, and homography estimation. A profile-marginal score marginalizes the nuisance inlier scale and selects an inlier partition, from which we estimate the scale that sets the LO loss width; this makes LO robust to an inlier scale specified too large, whereas one specified too small degrades selection itself. Refinement needs gradient from correspondences the seed currently rejects: optimizers that reweight from current residuals stay pinned to their seed, whereas methods with broad basins recover strongly perturbed seeds yet degrade accurate score-selected hypotheses, so basin size alone is insufficient to assess RANSAC LO. Joint half-quadratic optimization balances the two and is the most consistent strategy across model classes. An optimizer matched to the profile-marginal score, which never decreases it, does not reach the best accuracy, challenging the prescription that scoring and refinement objectives should match. Composed from these findings, our RANSAC reduces the median essential-matrix pose error of a state-of-the-art RANSAC on PhotoTourism from 2.23 degrees to 1.58 degrees with a correctly specified inlier scale and from 38 degrees to 6.2 degrees when it is grossly misspecified (128x too large).

This paper studies an inexact Halpern-accelerated generalized proximal point algorithm with an admissible positive semidefinite preconditioner $\mathcal{M}$ for solving maximal monotone inclusion problems. The proposed framework combines Halpern anchoring, resolvent inexactness, and relaxation over the full weight range $\rho\in(0,2]$, thereby covering both under-relaxed and over-relaxed proximal iterations within a unified scheme. We first establish convergence of the inexact resolvent sequence under conditions that allow general anchoring parameters and certain nonsummable tolerances. For anchoring parameters $\beta_{k}=1/(k+r)$ with $r\geq2$, we then derive explicit bounds on the squared fixed-point residual in the $\mathcal{M}$-seminorm. In particular, if the tolerances satisfy $\varepsilon_{k}=\mathcal{O}((k+1)^{-\alpha})$ with $\alpha>3/2$ for $0<\rho<2$ and $\alpha>2$ for $\rho=2$, these bounds yield an $\mathcal{O}(1/k^{2})$ convergence rate. The stronger decay condition on the inexactness tolerances at $\rho=2$ highlights a qualitative distinction between the endpoint and the interior regime in the inexact setting. Finally, we develop inexact accelerated versions of the preconditioned alternating direction method of multipliers (pADMM) and the preconditioned primal–dual hybrid gradient (PDHG) method based on this framework, derive inexactness criteria based on subproblem residuals, and establish a nonergodic $\mathcal{O}(1/k)$ KKT residual rate for their inexact iterates.

Credibility theory combines individual experience with portfolio information for insurance pricing, but classical formulations focus primarily on conditional means and expected premiums. We propose a Dirichlet mixed-membership model (DMMM) for multivariate distributional credibility. Policyholder risk is represented by a stable composition over latent risk classes: baseline characteristics determine its a priori assessment, while repeated multivariate experience progressively updates its a posteriori assessment. Support-specific expert distributions accommodate heterogeneous outcomes, with dependence induced through the shared latent structure. The resulting posterior predictive distribution can be written exactly as a convex combination of portfolio and experience components, extending the familiar credibility-factor structure beyond the conditional mean. We study the framework through a simulation experiment and a vehicle telematics application. The simulation shows that relatively simple experts can capture complex multivariate insurance distributions and learn policyholder-specific risk as experience accumulates. In the telematics application, recent claims and driving behaviour provide complementary information for future claim-frequency prediction, allowing similar policyholders to receive different experience-rated assessments. The DMMM therefore provides a flexible and interpretable framework for combining heterogeneous insurance experience while retaining the structure of classical credibility.

Market-regime models typically assume a finite set of discrete latent states. We examine whether high frequency limit-order-book dynamics exhibit distinct regime separation or apparent regimes result from discretising an underlying continuum, analysing deep limit-order book data for EURO STOXX 50 index futures across 987 clean trading days from 2022 to 2025 using 30, 45 and 60-second aggregation windows. Market states are represented by symmetric positive definite covariance matrices and analysed under Log-Euclidean and affine-invariant geometries; the resulting state cloud has an effective dimension of approximately 1.4. A causal anomaly layer filters statistical, calendar and rollover contamination, while label-free economic validation evaluates market-state separation independently of volatility-based proxy labels. A 17-method zoo spanning conventional and Riemannian representations, reinforced by full-cloud geometric certificates, consistently favours a continuum over a discrete regime structure. A dominant latent coordinate captures between 83.66% and 85.27% of variation in the covariance-state geometry and tracks VSTOXX without regime labels, remaining stable across years although its average level shifts with market conditions, indicating that conventional regimes are coarse quantisations. A divisive hierarchy spanning macro, macro-subregime, micro and micro-subregime tiers is evaluated using expanding-year walk-forward tests, with out-of-sample performance strongest at the macro tier while finer tiers fail to generalise; continuous fine-scale information nonetheless remains predictive, and its discretisation causes the performance loss. Forecasting performance peaks near ten minutes, reaching a maximum pooled R2_OOS of 0.7483. Overall, the latent stress coordinate defines a stable, low-dimensional geometric continuum that captures market stress and its predictive dynamics.