Papers

A geometry-based framework for sampling electromagnetic fields over surfaces is presented. The approach extends the shadow-area formulation of electromagnetic degrees of freedom (DoF) from a global mode count to a spatially resolved DoF density that determines the sampling distribution. The observation surface is partitioned into cells containing approximately one DoF, with the cell area determined by the DoF density and the cell shape determined by the projection of the DoF density vector onto the observation surface. In the far field, the source shadow area defines a directional DoF density over the observation sphere, while in the near field, the mutual-shadow density and a DoF density direction determine the local sampling geometry. The resulting method requires only the source-observation geometry and does not require an explicit singular-value decomposition of the propagation operator. Numerical results for several source and observation geometries demonstrate reconstruction performance close to that of operator-based sampling methods. A beamforming interpretation further shows that the sampling cells correspond to geometry-dependent beam regions, establishing a unified connection between DoF, sampling, and beamforming.

In upcoming integrated sensing and communications (ISAC) systems, security depends on whether an eavesdropper (Eve) can recover data symbols or infer sensing information from a common observation with a legitimate receiver. However, existing information and estimation measures do not quantify this dual risk under finite observation records. This letter introduces the novel metric of operational leakage, which signifies Eve's optimal probability, in the Bayesian sense, of reconstructing communication symbols or estimating a sensing parameter within prescribed distortions. By showing that prior adversarial success probability separates the effect of side information from that of the observation, we propose a conditional Laplace approximation for efficiently evaluating the proposed metric and jointly quantifying the eavesdropping risk to the sensing and communication functionalities. Numerical results showcase that the risk enabled by adversarial observation increases with Eve's signal-to-noise ratio (SNR), conventional metrics do not uniquely determine finite-record adversarial success probability, and that the proposed conditional Gaussian approximation closely follows numerical Bayes evaluation.

This paper summarizes the Interspeech2026 second Multilingual Conversational Speech Language Model (MLC-SLM) Challenge, which aims to advance the development of effective multilingual conversational speech language models. We describe the two challenge tasks: multilingual conversational speech diarization and recognition, and multilingual conversational speech understanding, together with the released real-world conversational speech dataset, evaluation protocols, and baseline systems. The challenge attracted 91 teams worldwide, with 704 valid leaderboard results and 14 technical reports across the two tasks. Based on the participating systems, we summarize representative approaches and distill practical insights into multilingual conversational speech recognition and understanding to support future research in the community.

Kalman-type filters are widely used for tracking dynamic systems, yet the confidence regions commonly derived from their estimated covariances can become unreliable under nonlinearities, non-Gaussian disturbances, and model mismatch. In this work, we develop a conformal prediction (CP) framework for equipping Kalman-type filters with statistically reliable confidence regions. Unlike CP applied to black-box estimators, our approach exploits the recursively estimated first- and second-order moments of the state posterior, which capture the time-varying uncertainty induced by the underlying dynamics. Based on these statistical features, we propose three complementary constructions. The first conformally calibrates Gaussian confidence regions induced by the filter moments. The second employs quantile regression to map the estimated moments into the boundaries of adaptive convex regions, which are subsequently calibrated. The third learns a Gaussian-mixture representation of the posterior and conformalizes the resulting density-based regions, enabling the characterization of multimodal and non-convex uncertainty sets. We establish finite- sample coverage guarantees for both sample-wise confidence, which controls miscoverage at individual time instances, and trajectory- wise confidence, which jointly covers the state sequence over a prescribed horizon. Numerical experiments across diverse linear and nonlinear dynamic systems demonstrate that the proposed methods attain the prescribed coverage while producing tight and informative confidence regions, and highlight the relative merits of the three constructions under different posterior characteristics.

Downlink multi-user multiple-input multiple-output (MIMO) precoding with limited channel state information (CSI) feedback in multi-cell systems is studied. Conventional codebook-based CSI feedback compresses the UE-specific physical channels, which leaves an inherent mismatch between the base-station (BS) precoders and the interference-aware UE combiners. To address this limitation, an effective CSI (ECSI) feedback framework is proposed in which UEs compute their linear combiners from precoded downlink pilots, construct corresponding post-combining effective channels, and report compact quantized representations using the same payload structure as in conventional schemes. The BS reconstructs the effective CSI and iteratively refines its precoders through over-the-air signaling without incurring additional downlink overhead. Analytical expressions are derived to characterize the effective-channel estimation error under both conventional CSI and ECSI feedback, explicitly capturing the impact of beam selection and rank truncation. Simulations demonstrate fast convergence within a few iterations and up to 30% average sum-rate improvement over conventional CSI-based precoding, with gains reaching 100% at the 10th percentile of per-stream rates in interference-limited conditions.

The deployment of unmanned aerial vehicles (UAVs) in low-altitude airspace requires accurate and timely radio maps for reliable communication and safe navigation. However, constructing such radio maps is challenging due to the prohibitive overhead of exhaustive measurements and the limited flight endurance of UAVs. To address this challenge, we propose an active learning framework based on flow matching for efficient low-altitude radio map construction from sparse measurements. We first analyze a plug-and-play (PnP) inference scheme with a flow-matching prior. By characterizing the late-stage refinement behavior through an ordinary differential equation (ODE), we theoretically show how the inference steps smoothly align with a continuous ODE flow to refine the map details. Recognizing that the early generative stages are largely noise-dominated, this insight motivates our proposed truncated flow matching plug-and-play (TFM-PnP) approach. TFM-PnP utilizes a spatial interpolation-based initialization to start the reconstruction from an intermediate flow time, thereby bypassing the inefficient early stages. We further use the generative diversity of flow matching to derive an uncertainty map to guide the UAV trajectory design. Specifically, we propose a weighted sampling approach to select a target location, followed by a Utility-Aware Path Search (UAPS) algorithm to design the corresponding UAV trajectories. Simulation results based on Sionna ray-tracing datasets show that the proposed framework outperforms the considered baselines, achieving more than 50% reduction in normalized mean squared error (NMSE).

This feasibility study presents a Hidden Markov Model (HMM)-based methodology for the Remaining Useful Life (RUL) estimation of rolling element bearings when the training vibration signals differ significantly from those of the target bearing. An AutoRegressive (AR) model is first employed to capture the machinery dynamics using vibration acceleration measurements from an initial operating period where all components including the considered bearing are still under healthy condition. The AR model is subsequently used to filter a newly acquired vibration signal from the faulty machinery, and typical Envelope Analysis is performed on the residual signal to identify fault-related repetition frequencies. The amplitudes of these frequencies, combined with statistical features of the AR residuals are fused through Principal Component Analysis to construct a sensitive to bearing degradation Condition Indicator (CI). Based on training data from the healthy state, a threshold is established, and a fault is declared when the CI exceeds this threshold. Once a fault is detected, its progression is modelled as a sequence of consecutive, distinct states, which are identified using K-Means clustering. A left-right HMM with continuous observation densities is estimated to represent the fault progression and thus predict the RUL. The HMM-based RUL estimation methodology is trained using vibration measurements from a limited number of two run-to-failure experiments with two nominally identical bearings, while RUL estimation performance is assessed on a third bearing of the same type. Despite the significant differences between the vibration data used for training and the life cycles of the training bearings with respect to the target bearing, the RUL estimation results indicate an adequate but conservative performance of the postulated HMM-based methodology.

Controlling the shape of a continuum soft robot typically requires an accurate dynamic model and actuation of all degrees of freedom. We show that regulating only the actuated coordinates, through collocated shape control, achieves provably stable convergence of those coordinates and, under an explicit compatibility condition, of the entire robot shape. While collocated control is a cornerstone of high-performance motion control in rigid robotics, extending this formulation to continuum soft robots has remained challenging due to the complexity of their dynamics. We present the first general framework for collocated control of continuum soft robots and derive a unified family of controllers, including PD, PID, PsatID, and their counterparts with compensation and cancellation components. The framework unifies existing approaches while introducing new controller designs. In particular, we develop three classes of PD and PID like regulators with local, semi-global, and global stability guarantees, and provide rigorous convergence analyses for each. Extensive experimental validation demonstrates the effectiveness of the proposed methods across different model discretizations and controller parameters. The resulting framework provides practical design guidelines for selecting and implementing controllers with known stability guarantees, without requiring a complete dynamic model of the robot

Action-chunked vision-language-action (VLA) policies improve inference efficiency, but limited feedback within committed action chunks can lead to accumulated execution errors. Residual adaptation can correct such deviations without retraining the VLA; however, existing corrections are typically optimized for reference-action consistency without explicitly considering their downstream consequences. To address this limitation, we present Cerebellum-Inspired Consequence-Aware Residual Governance (CereVLA), a unified framework that integrates lightweight residual refinement and predictive consequence evaluation into frozen VLA execution. Corrective actions are first generated by flow-based residual refinement, and their short- and interval-horizon consequences are then evaluated by a recurrent state-space model and a history-aware classifier. Residual corrections predicted to be unfavorable are selectively suppressed by a lightweight governor. Comparisons with state-of-the-art methods on LIBERO-10 and LIBERO-GOAL demonstrate the effectiveness of CereVLA. On SO-101, CereVLA increases task success from 57.5% to 90.0% and reduces mean control steps by 19.6% among successful trials, relative to the frozen SmolVLA baseline.

Distributed model predictive control (DMPC) often constructs both predictions and collision constraints from neighbor trajectories, so packet loss can remove both. We separate these roles: received trajectories drive Koopman-MPC, while local sensing and shelf geometry define a hard-constrained quadratic program (QP) that projects the applied input. Its radial demand is the least constant acceleration that keeps a supporting-plane clearance nonnegative throughout one zero-order-hold interval. Complementary pair rows recover the coupled demand without exchanging safety decisions. We give an intersample separation theorem under bounded snapshot and directional plant errors, an exact max-min test for simultaneous local feasibility, and a sensing-radius condition for switching interaction graphs. Anticipatory high-order rows may be relaxed for performance, but the finite-hold rows contain no safety slack. Matched eight-robot warehouse simulations use a frozen Koopman model, nonlinear drift, bounded inputs and speed, shelf constraints, a 120 ms control period, and packet dropout. The full controller is collision-free in 20/20 matched trials and reaches 160/160 robot goals; predictive Koopman-MPC without the final projection is collision-free in 1/20 trials. All 38,400 full-method hard-row sets pass the online feasibility test, and every local QP solves. Five-stream fleet sweeps are collision-free and hard-row feasible through 16 robots; the 20-robot boundary fails only after the online margin turns negative, while the reconstructed per-agent critical path remains below the sampling period. Bounded-sensing and differential-drive tests provide additional deployment stress.

3D reconstruction from satellite imagery is essential for large-scale topographic analysis, yet the lack of high-fidelity training datasets with accurate occlusion labels remains a primary bottleneck. Existing benchmarks, such as US3D and WHU-Stereo, face inherent challenges in spatio-temporal mismatch – environmental changes and shadow displacements between multi-view acquisitions – and provide ambiguous ground truth in occluded regions due to LiDAR sparsity. In this paper, we propose SatUnreal, a high-precision synthetic dataset designed to fundamentally overcome these limitations through an Unreal Engine-based simulation pipeline. SatUnreal provides 10,000 stereo pairs with high resolution (0.3m GSD) and is characterized by: (1) Physical Geometry Simulation, replicating realistic satellite orbits by systematically varying baselines and azimuths; (2) Spatio-temporal Consistency, eliminating temporal noise through fixed virtual environments; (3) Topographic Diversity, spanning dense urban canyons to low-texture natural terrains; and (4) Mathematical Label Integrity, utilizing a novel two-step linetrace algorithm to generate flawless occlusion masks. Experimental results using SOTA iterative models demonstrate that models trained exclusively on SatUnreal achieve superior zero-shot transfer performance on real-world benchmarks (US3D, WHU-Stereo) compared to those trained on real datasets. Our findings prove that physically accurate synthetic data provides a more effective supervisory signal for learning geometric features than complex real-world observations, establishing a new paradigm for Sim-to-Real transfer in Earth Observation. Code and dataset are available at https://github.com/jmp-Telepix/SatUnreal_A_High-Precision_Synthetic_Dataset_for_Satellite_Stereo_Matching_via_UnrealEngine

Achieving stable long-term neural decoding in invasive brain-machine interfaces (BMIs) remains challenging due to variations in recorded neural populations across sessions. Current latent alignment approaches may overlook task-dependent structure during cross-session adaptation. We propose Task-Conditioned Latent Alignment (TCLA), a framework that stabilizes neural decoding by learning a shared latent space. TCLA learns a low-dimensional source representation using neural reconstruction and continuous behavioral supervision. During target-session adaptation, the shared representation is fixed, while target neural activity is mapped into the source latent space by aligning source and target distributions separately for each task condition. We evaluated TCLA on seven nonhuman primate datasets spanning multiple tasks. In long-term cross-session evaluation, TCLA achieved a mean $R^2$ of $0.476\pm0.014$ with a negative $R^2$ failure rate of only 6.8%. Across 1,356 within-subject session pairs, TCLA achieved a mean $R^2$ of $0.371\pm0.009$ with a failure rate of 6.8%. Across 2,134 cross-subject session pairs, TCLA achieved a mean $R^2$ of $0.218\pm0.004$ with a failure rate of 12.9%, substantially better than those of the comparison methods. These results demonstrate that by preserving behaviorally relevant and task-dependent latent structure, TCLA improves the robustness of neural decoding across recording sessions and subjects. The source code is publicly available at \href{https://github.com/FAMD-CASIA/TCLA}{https://github.com/FAMD-CASIA/TCLA}.

Fault detection in rotating machinery under varying operating conditions remains a challenging problem. Existing approaches are mainly based either on non-parametric vibration signal representations, parametric deep learning models, or on parametric ARMA-type models. While deep learning approaches often lack transparency and require large training datasets, ARMA-type models rely on the assumption of a purely rational spectrum, which can be restrictive for rotating machinery vibration signals exhibiting mixed spectral characteristics. To address these limitations, a novel Harmonic Regression ARMA (HR-ARMA) model is presented for rotating machinery dynamics identification, which combines an explicit harmonic regression part for representing the dominant deterministic cyclical components with an ARMA part for the remaining broadband stochastic dynamics. Based on this model and within a Multiple Model framework, fault detection is achieved under varying operating conditions. The HR-ARMA modelling performance is validated through simulation and experimental studies on a single-stage gearbox, demonstrating accurate representation of both cyclical and broadband dynamics, improved parsimony and significantly higher fault detection performance compared with conventional ARMA models, achieving at least 98% True Positive Rate at 5% False Positive Rate in the experimental study.

Machine fault diagnosis from vibration requires learning from scarce labelled fault recordings while meeting the computational constraints of edge devices for local inference. We introduce DualRes, a compact oscillatory state-space model that combines two complementary spectral views of vibration, capturing rapid changes and fine frequency structure. Time-aligned views are processed by selective oscillatory memory, which learns how long to retain temporal patterns. The encoder contains 39,528 parameters. We evaluate supervised learning across six bearing datasets and a gearbox benchmark, with an additional gearbox pilot. Recording-level splits and explicit accounting of labelled duration distinguish data efficiency from repeated exposure to correlated samples. On the main gearbox benchmark, DualRes achieves state-of-the-art performance among the nine evaluated methods at six of seven label budgets. With about six labelled seconds per class, it improves macro-F1 by 16.1 percentage points over the next strongest comparator. On the same benchmark, DualRes achieves a 1.44-fold recording-level speedup and a 24.8-fold reduction in checkpoint storage relative to a selective state-space baseline under matched hardware and runtime conditions. Bearing results reveal task-dependent trade-offs. These findings support oscillatory memory as a compact approach to vibration diagnosis under limited labelled exposure.

Anomalies are rare, and anomalous data are often unavailable during development, making it difficult to determine which anomaly detection models and configurations will generalize to unseen anomalies. Recent approaches address this challenge by generating pseudo-anomalies and using bounds on the achievable area under the ROC curve (AUC) to select the optimal configuration from a finite set of candidates. Instead, we use the AUC bound as a differentiable, anomaly-free objective for directly optimizing continuous parameters of anomaly detection systems. We demonstrate this framework by optimizing ensemble weights and introducing a learnable score-rescaling mechanism that adapts pseudo-anomaly scores, enabling optimization beyond a predefined candidate set. Experiments across multiple datasets and embedding models show that AUC-bound optimization achieves significant performance gains over conventional model selection and prior development-set-based parameter selection. The results further show that direct optimization is less sensitive to the choice of pseudo-anomaly construction.

Multiuser transmission with multiple radio-frequency (RF) chains enables flexible spatial multiplexing but incurs increased hardware complexity and power consumption. A programmable metasurface (MTS) provides an alternative means of introducing spatial degrees of freedom with a single-RF-chain access point (AP). This paper proposes Semantic Prism, an MTS-enabled framework for concurrently transmitting independent semantic messages to multiple users. Semantic Prism jointly maps multiuser semantic representations to a common AP-symbol sequence and a corresponding sequence of discrete MTS configurations, enabling each user to recover its intended message from the received signal sequence. This mapping is realized by a learning-based modulator that takes both the multiuser semantic representations and channel state information (CSI) as inputs, while accounting for the discrete phase-control constraint of the MTS. To facilitate practical deployment, we further develop a field-channel fine-tuning method that enhances robustness to CSI uncertainty. We successfully implement Semantic Prism on a 400-element MTS prototype with 2-bit phase control operating at 3.5-GHz. Field tests show that Semantic Prism substantially improves multiuser semantic reconstruction over the considered baselines and enables a single-RF-chain AP to concurrently serve up to seven users.

We propose a sensor pairing method for online target tracking based on time-difference-of-arrival (TDOA) measurements through a sensor network. Existing pairing designs fix active sensor pairs in advance. In online tracking, however, adapting the pairing to the current target estimate maintains a favorable sensor-target geometry around it. We select $K$ pairs at each time step to maximize a Fisher information matrix (FIM) criterion under a communication-degree constraint. The original combinatorial problem is cast as a mixed-integer second-order cone program. We solve it at every time step to obtain a globally optimal pairing. Experiments under various noise settings show that the proposed method obtains the lowest tracking error among the methods under the same communication-degree constraint.

Open RAN (O-RAN) slicing xApps must adapt resource allocations to changing channel conditions and traffic demands while meeting service-level agreements (SLAs). Deep reinforcement learning can produce adaptive policies, but their allocation rules remain encoded in neural-network parameters. Our goal is to retain this adaptability while making the controller's decision logic directly inspectable and editable by operators. We use a large language model (LLM) to evolve slicing controllers as compact Python programs whose decision logic remains readable and editable after optimization. The LLM proposes and revises candidates offline, while a calibrated simulator scores them, and the selected decision module runs unchanged in the O-RAN control path. On the NSF POWDER 5G testbed, the evolved controller releases resources from a guaranteed slice whose throughput target becomes unattainable under a sustained channel fade, improving best-effort throughput from 158.2 to 228.6 Mbps, a 44.5% gain over the best static allocation. Since the controllers are readable source code, their behavior can be predicted from their equations, defects can be diagnosed by reading the code, and calibration errors can be corrected with one-line edits, reducing SLA misses from 79.9% to 2.2% in one case and more than doubling fitness in another. In a four-slice trace-driven simulation calibrated to the same testbed, evolutionary search achieves higher average evaluation scores than independent prompting at a matched proposal budget, with mean normalized gains on held-out traces of 16.3% for prompting alone, 32.1% for evolution from scratch, and 51.0% for evolution from a starting program.

Speaker diarization must solve two problems: counting how many speakers are present in a given conversation, and assigning speech to each one. This becomes harder in crowded conversations with five or more speakers. We evaluate how end-to-end neural diarization models count speakers, and find systematic under-counting in crowded recordings. The standard diarization error rate (DER) hides this failure, because it is duration-weighted and barely penalizes the dropped, low-activity speakers. We therefore also report the Jaccard error rate (JER) and explicit counting metrics. We propose a gated loss that couples speaker existence with frame activity. This loss can be computed in two ways, duration-weighted or speaker-weighted, and the speaker-weighted variant, used as a regularizer, reduces the under-count. On real-world crowded recordings our method clearly improves the counting metrics, lowering JER by about 6% relative and DER by about 9% relative, while leaving sparse recordings unharmed.

Robots deployed for competitive tasks must outmaneuver their opponents without sacrificing safety. Existing approaches, including safe reinforcement learning (RL), train a single policy to achieve task success and avoid failures simultaneously. This coupling can complicate training and leave the learned policy exploitable by deliberate attacks. We propose Safety to Competence (S2C), a two-stage RL framework that separates safety synthesis from competitive task learning. We formulate competitive interactions as safety-critical Markov games and prove that perfect filtering preserves policy non-exploitability when all players commit to safe maneuvers. S2C learns a robust safety filter via adversarial RL, embeds it in the environment during task policy training, and retains the same filter at deployment. In simulated touchdown games, S2C outperforms eight safe RL baselines, achieving the highest win rate and Elo rating, and the lowest exploitability. Hardware stress tests against a human opponent confirm S2C's competence.

Validating Autonomous Driving Systems (ADS) in simulation requires testing architectures that can discover rare, safety-critical failures while generating scenarios that are executable, diverse, and useful for downstream failure analysis. We introduce Teach-to-Crash, a closed-loop testing framework that combines a constrained ego-centric scenario representation, stagnation-aware search control, and a dual-LLM architecture for adaptive failure discovery. A high-reasoning Teacher LLM acts as an adaptive search controller, while a low-reasoning Student LLM emits simulator-executable scenarios in a strict JSON schema. The Teacher intervenes only when rolling collision rate and time-to-collision metrics stagnate, providing strategic guidance to redirect the search. In a CARLA case study with two experimental setups that vary the ego vehicle's speed policy, Teach-to-Crash achieves the highest Collision Hit Rate (90.79%), the shortest mean Time-to-Collision (18.31 s), and a competitive Collision Discovery Rate (136.21). PAFOT attains a higher mean CDR (179.44), but with substantially larger variance. Teach-to-Crash also yields the highest diversity (0.547) and, averaged across both setups on the CARLA Traffic Manager controller, the highest avoidability-based usefulness proxy (60.04%) among the compared methods. These results, within the evaluated CARLA scope, provide evidence that closed-loop dual-LLM reasoning can steer adversarial simulation-based testing over a constrained executable program space, generating failures that are frequent, structurally diverse, and assessed as more frequently avoidable.

Learning a sparse graph from scarce data is practically important but challenging. Motivated by the desirable combination of local sparsity and strong global connectivity exhibited by expander-like graphs, we propose spectral connectivity-regularized graph learning (SCoGL), a framework that incorporates a family of Laplacian spectral priors to explicitly promote global connectivity. Specifically, SCoGL augments a combinatorial-Laplacian-constrained graphical lasso (GLASSO) objective over a target adjacency matrix $\mathbf{W}$ with a general connectivity prior computed from Laplacian eigenvalues. We derive gradients for several representative connectivity priors and develop a projected gradient descent (PGD) algorithm with Armijo backtracking to efficiently optimize $\mathbf{W}$. Experiments show that the proposed SCoGL variants improve graph recovery and enhance downstream tasks such as graph signal denoising when signal observations are scarce.

Self-regulation together with mutual promoter competition provides a simple mechanism for bistability in gene-regulatory models. We study a two-gene system with regulatory terms of Hill exponent one, allowing distinct basal production rates and distinct degradation rates. Each gene product, when bound to its own promoter, may enhance or reduce production relative to the basal rate, while the two products compete through promoter occupancy. We show that the system has at least one and at most three equilibria in the positive quadrant. Exactly two positive equilibria can occur only if one nullcline intersection is degenerate; consequently, a configuration in which all positive nullcline intersections are transverse has either one or three positive equilibria. If there are exactly three distinct positive equilibria, then all three are automatically hyperbolic: the two outer equilibria are asymptotically stable nodes and the middle equilibrium is a saddle. Moreover, every positive solution converges to an equilibrium. Hence the positive quadrant is the disjoint union of the basins of attraction of the two stable nodes and the one-dimensional stable manifold of the saddle, yielding global bistability.

Self-supervised learning (SSL) for speech and audio has largely progressed along separate tracks: speech models emphasise time-domain prediction, whereas audio representation learning has focused on time-frequency patterns. This separation creates a compatibility gap, limiting cross-domain generalization. In this work, we introduce JASPER, Joint Audio and Speech Pre-trained Encoder Representations, a framework that augments speech-pretrained models with time-frequency objectives. Specifically, JASPER performs masked prediction of temporal and spectral targets over long audio segments, enabling spectro-temporal representation learning of speech and audio signals. The proposed method consistently outperforms multiple baselines and existing speech/audio encoders on diverse speech, audio, and music tasks, demonstrating the effectiveness of unified spectro-temporal modeling.

A neural network's layers can be treated as time steps of a state-space model, turning Bayesian training into a smoothing problem: a forward pass propagates Gaussian moments through the network, and a backward Rauch–Tung–Striebel pass updates the weight posteriors in closed form. Such methods learn from each observation in a single pass, in an uncertainty-aware manner, and without gradient-based iterations or replay, which makes them well suited for online adaptation and data-efficient learning. Existing smoothing-based methods, however, are restricted to diagonal covariances across activations, discarding correlations between neurons. We overcome this limitation via a cross-covariance identity that enables full-covariance propagation through a network's nonlinear activations. We derive a one-step-per-layer smoother that approximates as Gaussian only each layer's affine output, and that applies both to deterministic systems with noisy observations and to stochastic systems described by output statistics. We demonstrate this method in non-stationary classification, online dynamics learning, and policy adaptation of a vision-language-action model, and find that it is generally more accurate than other smoothing-based methods.

Voice conversion (VC) systems fall into two families: non-parametric embedding-space methods, which need no trained model but degrade on short target utterances, and spectrogram-based neural architectures, which achieve strong quality via multi-module pipelines with tens of millions of parameters. We propose kNN-FM-VC, a single conditional flow-matching network that learns to approximate the kNN-VC mapping between WavLM embedding distributions of source and target speakers, replacing explicit pointwise kNN matching with a neural regressor trained on kNN-generated pairs. The model is conditioned on the target speaker via cross-attention and FiLM, and trained under three Gaussian conditional paths (Schrödinger bridge, straight line, and constant-variance Gaussian tube), enabling few-step sampling. Unlike Phoneme Hallucinator, which uses an upsampling stage followed by kNN matching, our 13M-parameter model performs conversion with a single learned network and supports one-step inference. On LibriSpeech, the one-step Gaussian Bridge achieves lower WER and higher estimated speech quality than FreeVC and Phoneme Hallucinator. Relative to kNN and kDOT, it substantially reduces WER.

Movable antenna (MA) systems reconfigure the wireless channel by repositioning the antenna within a confined region. The position optimization behind this capability has been studied under the premise that the channel at every candidate position is known. However, acquiring that knowledge is costly: a position can be measured only after the antenna has moved there, which takes many slots at the speed of the actuator, and the channel drifts meanwhile. We formulate MA positioning as a non-stationary kernelized bandit with a reachability constraint, in which exploration, tracking, and the reward lost in transit are coupled through a single physical action. We propose MoveUCB, which addresses the reachability constraint through a global target search, a persistence rule that holds the target across slots, and a transit-aware movement cost term. We prove that MoveUCB attains sublinear dynamic regret under the standard variation-budget condition of non-stationary bandits alone, with the physical parameters entering only through constants. Simulations against a dynamic-programming oracle and movement-unaware baselines show that MoveUCB attains the lowest regret with about half the travel of the closest baseline that moves.

Converter terminal characteristics are central to both dynamic interactions with the network and steady-state power sharing in inverter-based microgrids. Conventional additive virtual impedance (VI) shapes these characteristics through a single virtual branch. This paper proposes virtual two-port control, which uses four coordinated transfer channels to reconstruct the terminal impedance. The connected network consequently observes the original converter through a virtually inserted twoport interface, with conventional virtual impedance recovered as a degenerate case. The proposed control transforms the original impedance through a matrix linear-fractional map. We also derive a necessary and sufficient condition for the reconstruction to be stable and proper. Among its various potential applications in microgrids, we develop three representative ones in detail: passivation with reduced control effort, uncertainty compression, and seriesshunt power-flow regulation. Simulations of VSG-controlled converters demonstrate the advantages of the proposed method over conventional VI in these applications.

Pneumonia is difficult to diagnose in older long-term care residents; multimorbidity and atypical presentations obscure signs, motivating operationally efficient objective testing. We analyzed multi-channel digital stethoscope recordings from 185 Japanese residents (73 pneumonia, 112 symptomatic without), using radiologist-confirmed chest X-rays and clinician diagnoses as supervisory signals that train convolutional neural networks, multimodal fusion, and channel-based variants with time-domain Grad-CAM interpretability. Models were evaluated with repeated patient-level cross-validation showing models with X-ray supervision outperformed clinician supervision (F1 0.729, accuracy 0.783 vs. F1 0.637, accuracy 0.711). Additionally, a three-channel selection protocol maintained performance (F1 0.736; accuracy 0.803), with two mid-thoracic sites ranking highest and Grad-CAM attention overlapping adventitious sounds. These findings indicate automated multi-channel lung-sound analysis can aid long-term care pneumonia diagnosis, with X-ray supervision being more reliable than clinical, and fewer channels preserving performance while lowering acquisition times.

Sparse communication in decentralized zeroth-order learning requires compatible peer-state coordinates. We characterize this one-hop condition and develop ZO-COSMO, coupling two-query estimation with average-preserving masked consensus using $q$ values per active link. Global supports serve all-neighbor mixing; matching updates require agreement only within each pair. We derive a sharp contraction-per-scalar bound within the matching class and convergence guarantees for the core and sparse-momentum updates. At fixed matching, exact moment identities characterize how shared directions preserve gradient-heterogeneity cancellation and redistribute estimation error and disagreement. Mechanism experiments cover unequal curvatures, noise, and sparse momentum. Further tests span $64$ synthetic agents and eight logical Qwen LoRA workers. At matched payload budgets, Qwen2-7B QNLI gains $3.65$ accuracy points over explicit-index Rand-$k$; edge-local updates gain $3.42$ and $2.53$ points over all-neighbor mixing on eight-worker complete and ring graphs. A matched-first-step ablation gives a $3.92$-point momentum benefit. Seed-aware and same-matching controls distinguish encoding, scheduling, and query correlation.