Papers

Roadside units (RSUs) in vehicle-to-infrastructure (V2I) corridors draw near-constant power while serving traffic that varies by more than an order of magnitude over the day, yet the 100 ms latency budget of 3GPP TS 22.185 leaves almost no room for the 107 ms restoration-and-service interval that follows a wake-up from sleep. This paper develops a duty-cycle model of latency-constrained RSU sleep that accounts for every wake-up cycle's transition energy, service energy, and violation exposure, and shows the latency constraint is equivalent to a budget: at most 14.3% of vehicles may arrive during slept hours. Scheduling becomes a budgeted hour-selection problem, formulated as a constrained Markov decision process and solved with a budget-repair portfolio that deterministically restores feasibility at the constraint boundary, making results nearly insensitive to state discretization. On five days of measured traffic from four signalized intersections in Kuwait City (762,050 vehicles), the repaired independent per-RSU policies reduce corridor energy by 39.5% relative to always-on operation at 99% latency compliance. A pre-specified evaluation of corridor cooperation finds no material benefit at any training-set size: the gain is zero at the two most strongly correlated pairs at the full sample (adjacent-intersection correlations rho = 0.947, 0.977), where upstream signals are statistically redundant with local observations, and marginal (0.8 percentage points) only at the weakest-coupled pair. The conditions under which cooperation can help, scheduling intervals below the inter-RSU transit time, are identified.

Neural compression is currently dominated by Nonlinear Transform Coding (NTC), which maps data to real-valued latents via continuous transforms. Despite its success, NTC suffers from train-test mismatch due to non-differentiable quantization, a “smoothness bias" inherent in continuous transforms that precludes optimality for certain sources, and a loss of “shaping gain" due to the complexity of including high-dimensional vector quantization. We propose SoftBinary Coding (SBC), an end-to-end learning paradigm that bypasses these limitations by using a stochastic binary latent space. In the spirit of vector quantization, SBC employs discrete representations and compresses them through a novel fast binary channel simulation scheme, for which we provide a proof of rate optimality. Experimental gains on information-theoretic sources provide both theoretical and practical closure to NTC's limitations, establishing discrete binary structures as a viable path toward reaching optimal rate–distortion bounds. Surprisingly, SBC also achieves state-of-the-art performance on vector quantization of i.i.d. sources, exceeding Trellis Coded Quantization of the Gaussian source.

This paper studies a probabilistic interpretation of input-to-state stability (ISS) bounds for estimation-error dynamics in continuous-time systems. We show that, if the aggregated disturbance satisfies a probabilistic envelope in an essential-supremum sense, then deterministic ISS bounds immediately induce high-probability error tubes. To make this interpretation constructive, we also provide explicit sufficient conditions based on quadratic Lyapunov inequalities and specialize them to positive and cooperative systems. The approach is illustrated on a positive compartment model with aggregated measurements, where ISS tubes are compared with nominal uncertainty bands produced by a Kalman–Bucy filter and by Gaussian and robust moving-horizon estimators. The examples show that ISS tubes provide a conservative but computationally light uncertainty baseline, while robust MHE is less sensitive to outlier contamination than Gaussian-based

Popular ASR test sets adopt inconsistent conventions for numbers, disfluencies, entities, and casing, while standard normalizers erase the format distinctions users care about. Current benchmarks therefore cannot measure whether a model follows user preferences for output style. We introduce PreferenceASR, a test set evaluating ASR systems on their ability to follow natural-language preference instructions across four categories: normalization, entities, disfluencies, and case. Built from seven open-source corpora via a two-stage LLM-assisted pipeline with human verification, it is evaluated with a preference-aware normalizer that selectively skips steps matching the active instruction. Benchmarking four models shows rankings shift across preference types, exposing quality differences traditional evaluation obscures. We publicly release the dataset.

Variable frame rate (VFR) coding has recently emerged in neural speech codecs, allocating fewer frames to redundant regions and more frames to rapidly changing speech. VFR must transmit side information about retained time steps, but prior gains are either not rigorously addressed or often minor once these overhead bits are included in total bitrate. We present Dynamic Token Masking (DTM)-Codec, a neural speech codec that demonstrates clear gains over fixed-frame-rate baselines under a strict matched-total-bitrate protocol. DTM keeps selected encoder tokens, fills masked positions with a learned <MASK> embedding, and transmits a binary keep-mask for position-aware decoding. We further introduce Path Length Equalization (PLE), a linear-time boundary selector for VFR coding that yields well-spread adaptive segments with negligible overhead. Across operating points, DTM-Codec broadly improves reconstruction quality and intelligibility over fixed-frame-rate baselines.

Noise-robust bandwidth expansion aims to reconstruct high-fidelity wideband speech from noisy low-resolution inputs. While flow matching has shown strong performance in speech generation, accurately recovering clean speech from noisy inputs remains challenging due to the ambiguity of velocity estimation under noise. In this work, we propose VeRe-Flow, a clean-guided flow matching framework that introduces multi-level clean supervision to guide the generative process toward clean speech. At the velocity level, we introduce velocity contrastive regularization, which attracts the predicted velocity toward the clean trajectory while repelling it from noisy trajectories. At the representation level, we incorporate representation alignment that aligns intermediate features with clean self-supervised learning representations. The results demonstrate that the proposed method achieves the lowest LSD and highest DNSMOS OVRL among all baselines, and the highest MOS among generative baselines.

Heterogeneous and in particular multi-altitude low Earth orbit (LEO) satellite constellations exhibit complex spatial and temporal structures, which require new modeling tools for their performance analysis. In this paper, we develop an orbit-aware stochastic geometry framework modeling today's LEO satellites on various orbits and various altitudes. In particular, we characterize such a system as the superposition of multiple Walker point processes and formulate it as a dynamical system determined by an initial condition and the rotation speeds of satellites and Earth. We show that when the speeds are rationally commensurable, the proposed satellite system is periodic. Then, we show that the system is ergodic when the speeds are rationally independent, establishing a theoretical link between time averages of the system and the expectation of it under the invariant measure. We derive the nearest-satellite distance distribution of a typical receiver at a given latitude and analyze the signal to interference-plus-noise ratio (SINR) coverage probability of the typical receiver. We then derive the ergodic throughput of the downlink communication to the typical receiver. Overall, the proposed framework offers a rigorous and tractable tool for analyzing downlink performance in Walker-type heterogeneous LEO satellite networks.

This paper addresses the problem of achieving both coarse and precise privacy in state estimation. Coarse privacy forces the eavesdropper's total mean-square error (MSE) to infinity, but errors along certain confidential directions may remain bounded. This motivates precise privacy, which additionally drives the MSE along prescribed directions to infinity. For coarse privacy, an analytical transformation is established, preserving the user's optimality and driving the eavesdropper's total MSE to infinity at a polynomial-exponential rate. A stochastic intermittent encryption scheme is further developed, and an explicit lower bound on the encryption probability is derived to guarantee divergence. For precise privacy, by analyzing the behavior of the Riccati equation on the unobservable subspace, we prove that the eavesdropper's directional MSE becomes unbounded if and only if the direction's unstable component lies outside the observable subspace. Finally, a systematic method is proposed to exclude target vectors from the observable subspace, forcing the directional MSE to infinity.

Operating envelopes (OEs) are increasingly used to allocate limits to distributed energy resources (DERs) while maintaining secure distribution network operation. In unbalanced low-voltage feeders, OE calculation based only on voltage magnitude and thermal constraints can yield overly optimistic limits because power quality constraints such as voltage unbalance are neglected. This paper proposes a three-phase unbalanced AC optimal power flow framework for computing coupled P–Q OEs with explicit voltage unbalance factor (VUF) constraints. In addition, two fairness mechanisms for allocating the available P–Q flexibility across multiple PV units are embedded and compared: (i) network-weighted proportional fairness and (ii) lexicographic max–min fairness. Case studies on unbalanced test feeders illustrate how VUF constraints reshape the P–Q feasible region and the impact of power quality-constrained operation. The comparison highlights the trade-off between the efficiency, equity, and practicality of fairness allocation methods.

The growing demands for higher throughput and cost-efficient wireless communications drive the need for receivers that are both simple to deploy and robust to hardware impairments and nonlinear environments. While classical model-based receivers and recently proposed deep neural network ( DNN) architectures provide complementary benefits, they either rely on simplified linear Gaussian assumptions, require considerable computational resources, or are tailored for a given setting and modulation. In this work, we propose a compact and modular DNN augmentation that universally refines the soft outputs of existing receivers (model-based or data-driven), addressing two distinct operating regimes: structurally incomplete soft information arising from reduced-complexity detectors, and degraded soft outputs caused by hardware impairments and synchronization errors. A key property of the proposed framework is its task-agnostic nature: operating without any knowledge of the specific source of unreliability, it produces well-calibrated log-likelihood ratios (LLRs) suitable for channel decoding. Our design leverages an element-wise scaled convolutional neural network tailored to perform learned interference cancellation across users and neighboring subcarriers, combined with a training algorithm that encourages accurate LLR s for soft channel decoding. Numerical results demonstrate that the proposed augmentation consistently improves diverse receiver algorithms in challenging channel conditions while incurring minimal overhead.

Reliable event detection underpins induced-seismicity monitoring for Carbon dioxide Capture and Storage (CCS) and geothermal operations, distributed acoustic sensing (DAS), and industrial condition monitoring. In each setting a detector must stay reliable both when sensors fail and when the signal is buried in noise. These two failure modes are routinely conflated, and architectural complexity is often credited with robustness it may not deserve. We assemble a unified binary event-detection benchmark from three physically distinct real sources – Hi-net seismic waveforms, Utah FORGE 2024 borehole DAS, and MAFAULDA industrial vibration – each mapped to a common 8-channel, 256-sample representation, and evaluate a fault-tolerant detector (CEPHALON) trained with per-sample sensor-dropout against standard detectors (a 1D convolutional network, a temporal convolutional network, and a compact Transformer) trained with an identical recipe. On clean data every model is near-perfect (AUC ~ 0.99). Under progressive sensor loss, simple models with sensor-dropout are already robust and CEPHALON holds no advantage. Under additive noise, however, CEPHALON degrades far more gracefully: at -2.5 dB its overall AUC is 0.939 versus 0.532-0.572 for the convolutional baselines. Same-architecture ablations isolate the cause: disabling internal redundancy at inference reduces the low-SNR advantage only modestly, whereas removing sensor-dropout training collapses it (0.899 to 0.603 at -5 dB). The training recipe is therefore the dominant cause and parallel redundancy only secondary. We release a complete, numbered, reproducible pipeline so that every figure can be regenerated.

We propose a robust Extended Kalman Filter (EKF) architecture for land navigation using an array of hundreds of low-cost micro-electromechanical systems (MEMS) inertial sensors. The main challenges in this setting are bursty sensor-specific bias errors, bias drift, and the need to aggregate many inertial measurements without increasing the computational burden of the navigation filter. To address these challenges, we introduce Robust Inertial Sensor Array Fusion (RISAF), a pre-filtering framework that combines dynamic percentile gating with real-time bias tracking before the EKF prediction step. The proposed aggregation suppresses anomalous sensor readings and compensates for individual sensor drift while preserving the vehicle-level kinematic signal. Because the resulting fused inertial measurements are passed to a standard EKF, the navigation filter retains a minimal state vector and supports real-time execution. We evaluate RISAF through extensive simulations and real-world field tests in GNSS-denied environments, with the data provided as supplementary material. Compared with a baseline that averages the sensor readings, RISAF achieves substantially improved azimuth accuracy and reduced drift accumulation. These results demonstrate that robust fusion of large MEMS inertial arrays can bridge a substantial part of the gap between cost-effective hardware and tactical-grade inertial navigation performance.

This letter addresses long-term fairness in uplink non-orthogonal multiple access integrated sensing and communication (NOMA-ISAC) systems. Existing resource allocation schemes that maximize instantaneous sum rate often favor strong users, leaving historically underserved users with poor long-term throughput. We propose PF-JUGPA, a proportional-fair scheduling based joint user grouping and power allocation method. PF-JUGPA first pre-selects users via a PF metric combining instantaneous rate proxies and historical averages, then performs fairness-aware grouping and power allocation by maximizing a weighted sum rate whose weights are inversely proportional to historical service rates. Simulation results show that PF-JUGPA significantly improves the Jain fairness index and weak-user average rates with only a modest sum-rate loss compared to sum-rate-oriented and round-robin baselines. The findings confirm that embedding long-term service history into both scheduling and resource allocation yields an effective throughput–fairness–sensing tradeoff in uplink NOMA-ISAC.

The non-terrestrial network (NTN) architecture has gained significant interest from the academia owing to its versatility and the ability to provide worldwide service. To achieve extremely high data rates in NTNs, as intended in the sixth-generation (6G) communication systems, millimeter wave (mmWave) and terahertz (THz) frequencies can be considered, enabling substantial bandwidth and data transmission capacity, which makes them highly suitable for NTN applications. However, these high-frequency signals suffer from significant propagation challenges, including atmospheric attenuation, pointing errors, and various environmental effects. Therefore, a comprehensive link budget analysis is essential to accurately assess the feasibility of mmWave/THz-based NTN systems. Existing studies in the literature often fail to fully capture certain frequency-, altitude-, and direction-dependent effects observed in mmWave/THz transmission or possible communication scenarios within the NTN architecture. In particular, while most prior works primarily focus on free-space loss or atmospheric attenuation, this study adopts a much more comprehensive approach. In this work, a detailed link budget analysis is conducted for mmWave/THz NTNs, considering free-space loss, atmospheric absorption, weather-induced effects, ionospheric disturbances, polarization mismatches, feeder losses, antenna and circuitry constraints, fading, pointing errors, and non-white noise characteristics. The results have revealed that the multi-layer structure of the NTN architecture can help reducing the excessive loss levels to a certain level that can be tolerated by high-gain directional antennas/arrays, providing multi-gigabit links and making mmWave/THz NTNs feasible for 6G communication systems.

Video-based Point Cloud Compression (V-PCC) encodes volumetric data by projecting 3D geometry and texture onto 2D video frames. To prevent spatial distortion and color bleeding during 3D reconstruction, this process requires 10-bit color depth and 4:2:2 chroma subsampling, rather than the standard 8-bit 4:2:0 format. Additionally, capturing high-density dynamic point clouds requires demanding encoding parameters, such as 8K resolution at framerates up to 120 fps. Historically, the lack of 4:2:2 chroma support in older GPU hardware encoders restricted real-time V-PCC to custom Application-Specific Integrated Circuits (ASICs). However, the recent introduction of NVIDIA's Blackwell GPU architecture, featuring on-chip hardware encoders with 10-bit 4:2:2 support, presents an opportunity to shift this workload to general-purpose hardware. This paper investigates the feasibility of such an approach. Using a commercially available Blackwell GPU equipped with four parallel on-die hardware encoders as a testbed, we evaluate the throughput, rate-distortion (RD) performance, and power consumption of 8K 10-bit 4:2:2 HEVC across various Split-Frame Encoding (SFE) configurations. Our results demonstrate that 4-way SFE achieves an encoding throughput of 122 fps, successfully meeting the strict real-time constraints of high-density V-PCC. Although the inability to exploit spatial redundancies across slice boundaries results in a BD-Rate penalty of up to 5%, the measured throughput and power efficiency establish standard, commercial off-the-shelf GPUs as a highly viable baseline for real-time volumetric video streaming.

Adaptive video streaming is a widely used technique for delivering video content over the internet. One of the key challenges is determining the optimal encoding settings for each video, which can vary significantly based on its content and characteristics. In this paper, we propose Compression Echo Contrastive Learning (CECL), a novel self-supervised learning framework for clustering videos based on their encoding complexity. Our method leverages the response of a video to compression - the Compression Echo - as a supervisory signal, allowing the model to capture underlying encoding characteristics during pretraining. We conduct extensive experiments to demonstrate the effectiveness of our learned representations for the downstream task of clustering videos by their encoding complexity. Our results show that CECL improves upon existing state-of-the-art visual encoders and delivers strong bitrate and quality savings against the fixed bitrate ladder.

Self-supervised learning (SSL) currently drives state-of-the-art performance in no-reference image quality assessment (NR-IQA). However, standard SSL pipelines uniformly apply synthetic distortions across the entire image field, which can limit their sensitivity to spatially localized and co-occurring degradations encountered in real-world content. In this work, we empirically expose this representational blind spot across existing state-of-the-art encoders, demonstrating their reduced sensitivity to spatially bounded image degradations. To bridge this gap, we introduce Spatial Localized Image Degradation Embeddings for Image Quality Assessment (SLIDE-IQA). SLIDE-IQA employs a dual-branch Vision Transformer framework that injects spatially bounded degradations into a contrastive pretraining objective. To handle the spatial complexity of these degradations, we introduce a Threshold-Bounded Exclusion Mechanism, a representational design choice that resolves structural conflicts arising from spatially localized distortions to ensure the latent space respects both degradation type and spatial scale. Finally, we show that SLIDE-IQA's synthetic-only pretraining significantly improves sensitivity to localized distortions, while achieving competitive performance on NR-IQA benchmarks against existing SSL NR-IQA models.

Post-hurricane damage assessment and repair scheduling can require computationally intensive simulation and optimization. This paper presents an integrated two-stage deep-learning tool for rapid damaged-line identification and repair-schedule computation. An available offline synthetic dataset for the IEEE 9500-node test feeder contains 1,700 hurricane scenarios with exposure features, grid metadata, fragility parameters, OpenDSS outputs, damaged-line labels, and Adaptive Large Neighborhood Search reference schedules. Stage 1 benchmarks MLP, ResMLP, and GraphSAGE, while Stage 2 compares MLP, DeepSets, and Set Transformer. The selected ResMLP-Set Transformer pipeline propagates Stage 1 errors into Stage 2 and achieves a damaged-job F1-score of 0.920, pairwise order agreement of 0.854, and start- and end-time mean absolute errors of 4.349 min and 4.486 min, respectively. The tool provides rapid initial repair-log decision support for new hurricane cases.

Speech-to-speech dialogue models increasingly depend on prosody and interactional nuance to convey social intent, yet benchmarks for these cues remain limited. We introduce StanceBench, a benchmark for measuring interpersonal stance in conversational speech and evaluating audio-capable LLMs as automated judges. Using the Seamless Interaction corpus, StanceBench (1) specifies 9 stance dimensions via role-prompt poles, (2) standardizes single-speaker and interaction-based evaluations, and (3) reports LLM-as-a-judge robustness, bias, and stance inference. Across evaluated stances, empathy and politeness are the easiest. Warmth and assertiveness are moderately separable with positivity skew/asymmetry. Honesty is the hardest and shows high prompt order bias, consistent with needing cross-turn evidence. Attentiveness is separable but aligns weakly with humans. Interaction stances are more context-sensitive, with threshold gaps and high variance, especially conflict regulation.

The goal of this exposition is twofold. First, it surveys the current status of the synthesis of rational multivariable ($n$-dimensional) passive linear shift-invariant bounded-real (or positive-real) transfer functions as scattering (or immittance) matrices of networks composed of a finite number of $n$ types of inductive and capacitive elements together with memoryless reciprocal and nonreciprocal components, such as transformers and gyrators. The theory seeks to extend the classical $1$-D network synthesis framework to the multidimensional ($n$-D) setting. Second, the exposition examines the fundamental connection between $n$-D network synthesis and Hilbert's 17th problem concerning the representation of positive polynomials as sums of squares. This connection is shown to be the principal mathematical obstacle to extending many classical synthesis results to higher dimensions. Both lossless and dissipative transfer functions are considered. While every $2$-D lossless transfer function can be synthesized, the classical procedure of embedding a dissipative transfer function into a lossless one by solving an associated matrix dilation problem can be executed via spectral factorization only partially in $2$-D and fails in general for $n>2$. Moreover, for $n>2$, even lossless transfer functions are not, in general, synthesizable, except for certain low-degree stable all-pass functions. These limitations ultimately stem from Hilbert's 17th problem referred to above. The exposition also reviews failure of several other classical $1$-D synthesis techniques, including matrix-factorization methods, thereby providing a comprehensive account of the current state of $n$-D network synthesis. Depending on the context, both continuous and discrete time formulations are employed; for the class of linear shift-invariant systems considered here, this change of setting entails no loss of generality.

Deep learning models for diabetic foot ulcer (DFU) segmentation routinely report high accuracy, but they are almost always trained and tested on the same dataset, leaving their behaviour on data from a different clinical source largely unmeasured. We benchmark three representative segmentation architectures – U-Net and DeepLabV3+ (convolutional) and SegFormer-B2 (Transformer) – under an identical, leakage-screened protocol: training on the combined FUSeg/AZH wound data and evaluating, without fine-tuning, on two independent external datasets (DFUC2022 and Medetec). All models achieve strong in-domain performance (Dice 0.80–0.83) but degrade substantially across datasets. The degradation is, however, architecture-dependent: SegFormer-B2 generalizes best on both external sets (DFUC2022 Dice 0.557, Medetec Dice 0.786), outperforming both convolutional models, while the more complex DeepLabV3+ generalizes worse than the simpler U-Net. Per-image failure analysis on 2,160 images across both external test sets confirms that SegFormer-B2 produces the fewest catastrophic failures on DFUC2022 (31.1%), compared with U-Net (38.5%) and DeepLabV3+ (43.0%). The consistent ranking across two independent external sources, confirmed by Wilcoxon signed-rank tests (p < 0.001 on both datasets), indicates that architecture family, not model complexity, drives cross-hospital generalization.

Recent advances in language–audio retrieval have been largely driven by contrastive dual-encoder architectures that align audio and text in a shared embedding space. While effective, existing retrieval embeddings are primarily optimized for audio–caption matching, limiting their ability to support diverse retrieval objectives and controllable retrieval behaviors. We present ALM2Vec, a universal audio embedding framework derived from pretrained large audio–language models (LALMs). By transferring the audio understanding, instruction-following, and reasoning capabilities acquired through large-scale multimodal training, ALM2Vec learns a unified embedding space for retrieval across audio domains and task types. Beyond conventional text–audio retrieval, ALM2Vec incorporates natural-language instructions into the embedding process, enabling instruction-aware retrieval for scenarios such as audio question answering and aspect-conditioned retrieval. Experimental results show that ALM2Vec achieves competitive performance on standard audio and speech retrieval benchmarks while exhibiting promising compositional and controllable retrieval capabilities, highlighting its potential as a unified audio embedding model for retrieval across domains, tasks, and user intents.

A growing number of techniques leverage the spatial structures that underlie many real-world datasets. Despite these advances, the complementary task of estimating spatial structures and understanding their role within these techniques has often been overlooked. In neurophysiological data analysis specifically, numerous methods exist to estimate brain connectivity, but most are not explicitly model-based, dynamic, multivariate, or directed. To address these limitations, we previously introduced noise-driven heat modelling on graphs for neurophysiological connectivity estimation. In this study, we extend this framework by relaxing earlier noise assumptions and adding regularisation to improve robustness. We also develop a simulation procedure to characterise and evaluate our technique in a controlled setting. Finally, we demonstrate that the technique is able to capture meaningful spatial structure across two experiments, each using two real-world datasets. The explicit model formulation of our connectivity estimator has the potential to improve the interpretability of graph-based techniques across a wide range of applications. The code implementing our method is available at https://github.com/sgoerttler/Heat_Connectivity.

The carbonized papyri from Herculaneum preserve the only large-scale library to survive from classical antiquity, but many unopened rolls remain unread because physical opening risks irreversible damage. X-ray computed microtomography ($\mu$CT) and virtual unwrapping offer a non-invasive route to their texts, yet previous work on sealed Herculaneum scrolls has recovered only localized readings or limited surface regions. Here, using high-resolution phase-contrast $\mu$CT acquired on the BM18 beamline at the European Synchrotron Radiation Facility (ESRF), together with improved computational unrolling and machine learning, we achieve the complete virtual unwrapping and reading of PHerc. 1667 under explicit coverage and papyrological-review criteria. This makes PHerc. 1667 the first Herculaneum papyrus to be fully digitally unrolled and read for extended scholarly study without physical opening. In PHerc. Paris 4, the optimized scan protocol makes ink directly visible in the tomographic volume, allowing three-dimensional ink segmentation and independent validation of surface-conditioned ink recovery. In PHerc. 139, we recover title and author-attribution evidence identifying the scroll as Philodemus, On Gods, Book 8. These results move virtual unwrapping of the Herculaneum scrolls beyond isolated demonstrations towards a scalable framework for systematic recovery of the still-unopened library.

This paper proposes a divergence-based safety measure for large language models (LLMs) under embedding-input attacks. The proposed measure quantifies the worst-case Kullback–Leibler divergence between the clean and attacked LLMs' output distributions, subject to a stealthiness constraint. This constraint is constructed by leveraging the equivalence between transformer attention used in LLMs and rational inattention modeling human decision-making. We analyze the proposed divergence-based safety measure by investigating perfectly undetectable attacks and deriving its upper bound through a Bregman-divergence argument. The proposed safety measure is applied to two pretrained causal language models, GPT-2 and GPT-Neo-125M, to show nontrivial output-distribution shifts, illustrating that the measure can distinguish model-level safety profiles.

This paper studies resilient control for cyber-physical systems operating under hidden degraded or compromised modes. We formulate hidden-mode detection and belief-dependent control as a game between two decision makers with different objectives: the detector seeks informative belief updates, while the controller seeks regulation performance. This objective mismatch shows why the usual separation intuition between detector design and controller design may fail, leading to a performance-reversal phenomenon induced by the resilience of the controller. For a two-mode linear Gaussian system, we theoretically characterize this phenomenon by linking the resilience margin to the log-likelihood evidence. The analysis shows that a well-performing controller with a large resilience margin can suppress mode-dependent information and slow belief adaptation, which in turn degrades the control performance. The resilience-induced blindness phenomenon and its partial mitigation through increased detector sensitivity are illustrated in numerical simulations.

High-altitude platform station (HAPS) massive multiple-input multiple-output (MIMO) systems are expected to support wide-area, low-latency, and energy-efficient connectivity in future non-terrestrial networks. However, Doppler-induced channel aging, finite-rate feedback quantization, packet loss, and estimation noise impair transmitter-side channel state information (CSI), making robust downlink beamforming challenging. In HAPS channels, these impairments are strongly structured by elevation-dependent Rician propagation and line-of-sight (LoS)-dominant geometry, whereas conventional robust beamforming methods often rely on generic uncertainty models and computationally intensive optimization. This paper develops a physics-informed uncertainty-aware beamforming framework for HAPS massive MIMO systems under imperfect CSI. First, a geometry-aware channel and feedback-impairment model is developed, where CSI errors due to aging, quantization, packet loss, and noise are represented through tangent-space ellipsoidal uncertainty sets. Second, a physics-informed variational autoencoder (VAE) exploits the LoS-dominant steering manifold to enhance channel direction information and propagate learned uncertainty through unit-sphere projection. Third, the learned uncertainty representation is embedded into a robust energy-efficiency maximization formulation with probabilistic QoS awareness. To enable scalable online operation, the resulting beamforming policy is approximated using a multi-agent deterministic policy gradient framework with centralized training, decentralized execution, and differentiable power projection. Simulation results show that the proposed framework improves energy efficiency, SINR robustness, outage reliability, convergence behavior, and online runtime compared with imperfect-CSI, SDR-based, and no-VAE baselines.

A lateral velocity estimation scheme whose stability can be analytically derived (rather than empirically demonstrated through cut-and-try) is attempted. The designed adaptive sliding mode observer shows robust performance under a wide variety of maneuvers/ environments, including the more challenging slow J-turn on low mu.

Single mass-spring-damper models of vocal folds have been effective in simulating vocal fold vibrations without added complexity. However, single-degree-of-freedom models cannot sustain oscillation in the presence of structural damping unless source-tract interaction is considered. Moreover, existing lumped models struggle to accurately simulate vocal fold closure during phonation. This study aims to develop a reliable and simplified single-degree-of-freedom model of phonation that can simulate sustained oscillation in a damped system without incorporating a vocal tract model. Additionally, the proposed model maintains vocal fold closure in a manner consistent with the physics of phonation, addressing a longstanding challenge in existing lumped models. High-speed videoendoscopy (HSV) data from four normophonic subjects producing sustained vowel /i/ were used to extract glottal area waveforms (GAWs) via deep learning-based image segmentation for particle swarm optimization of the model parameters. An additional resistance force was incorporated to compensate for flow separation and generate the force imbalance required for sustained oscillation. An external structural force was also added during closure to sustain the closed phase. The 4th-order Runge-Kutta method was used to solve the governing equations with enhanced numerical stability and accuracy. The model parameters were optimized for individual subjects, resulting in normalized errors below 3% between experimental and simulated GAWs. The proposed model accurately reproduced subject-specific vocal fold vibrations and vocal fold closure in agreement with experimental data. Overall, the proposed model provides a computationally efficient framework for simulating sustained phonation without requiring complex source-tract coupling while capturing the key biomechanical and aerodynamic mechanisms of phonation.

The DC optimal power flow is widely used in power system operations because of its computational efficiency and scalability. However, DC dispatches are not guaranteed to satisfy the nonlinear AC power-flow equations or associated operational limits. This paper develops a parameterized, differentiable AC power-flow restoration method for mapping DC dispatches to AC-consistent operating points. The method incorporates distributed slack for active-power balancing and PV/PQ switching for reactive-power regulation, both implemented using smooth differentiable surrogates with tunable parameters, including slack participation factors, voltage setpoints, and regulation steepness. These parameters are trained offline by differentiating through the AC restoration equations using the implicit function theorem. Once trained, the optimized parameters are fixed and used directly during AC power-flow recovery from DC dispatches. The approach is evaluated on IEEE, ACTIVSg, and PEGASE test systems using setpoints computed by standard DC optimal power flow. Results show that the optimized restoration method improves AC feasibility recovery across various systems relative to conventional single-slack AC power-flow recovery. On the 9,241-bus case, the optimized method improves cost difference by 80% relative to the conventional recovery baseline and improves solving time relative to ACOPF by 75%.