Papers

Scaling spoken language modeling requires speech tokens that are both efficient and universal. Recent work has proposed syllables as promising speech tokens at low temporal resolution, but existing models are constrained to English and fail to capture sufficient acoustic detail. To address this gap, we present Sylber 2.0, a self-supervised framework for coding speech at the syllable level that enables efficient temporal compression and high-fidelity reconstruction. Sylber 2.0 achieves a very low token frequency around 5 Hz, while retaining both linguistic and acoustic detail across multiple languages and expressive styles. Experiments show that it performs on par with previous models operating on high-frequency baselines. Furthermore, Sylber 2.0 enables efficient TTS modeling which can generate speech with competitive intelligibility and quality with SOTA models using only 72M parameters. Moreover, the universality of Sylber 2.0 provides more effective features for low resource ASR than previous speech coding frameworks. In sum, we establish an effective syllable-level abstraction for general spoken language.

The extraction of signals from noise is a common problem in all areas of science and engineering. A particularly useful version is that of forecasting: determining a causal filter that estimates a future value of a hidden process from past observations. Current techniques for deriving the filter require that the noise be well described by rational power spectra. However, scale-free noises, whose spectra scale as a non-integer power of frequency, are ubiquitous in practice. We establish a method, together with performance guarantees, that solves the forecasting problem in the presence of scale-free noise. Via the duality between estimation and control, our technique can be used to design control for distributed systems. These results will have wide-ranging applications in neuroscience, finance, fluid dynamics, and quantum measurements.

LLM-based and agent-based synthetic personas are increasingly used in design and product decision-making, yet prior work shows that prompt-based personas often produce persuasive but unverifiable responses that obscure their evidentiary basis. We present PersonaCite, an agentic system that reframes AI personas as evidence-bounded research instruments through retrieval-augmented interaction. Unlike prior approaches that rely on prompt-based roleplaying, PersonaCite retrieves actual voice-of-customer artifacts during each conversation turn, constrains responses to retrieved evidence, explicitly abstains when evidence is missing, and provides response-level source attribution. Through semi-structured interviews and deployment study with 14 industry experts, we identify preliminary findings on perceived benefits, validity concerns, and design tensions, and propose Persona Provenance Cards as a documentation pattern for responsible AI persona use in human-centered design workflows.

Due to their low-complexity and energy-efficiency, unified simultaneous wireless information and power transfer (U-SWIPT) receivers are especially suitable for low-power Internet of Things (IoT) applications. Towards accurately modeling practical operating conditions, in this study, we provide a unified transient framework for a dual-diode U-SWIPT that jointly accounts for diode nonlinearity and capacitor-induced memory effects. The proposed model accurately describes the inherent time dependence of the rectifier, highlighting its fundamental impact on both energy harvesting (EH) and information decoding (ID) processes. Based on the provided memory-aware model, we design a low-complexity adaptive detector that learns the nonlinear state transition dynamics and performs decision-directed detection with linear complexity. The proposed detection scheme approaches maximum likelihood sequence detection (MLSD) performance in memory-dominated regimes, while avoiding the exponential search required by classical sequence detection. Overall, these results demonstrate that properly exploiting rectifier memory provides a better tradeoff between data rate and reliability for U-SWIPT receivers.

We propose a brain-informed speech separation method for cochlear implants (CIs) that uses electroencephalography (EEG)-derived attention cues to guide enhancement toward the attended speaker. An attention-guided network fuses audio mixtures with EEG features through a lightweight fusion layer, producing attended-source electrodograms for CI stimulation while resolving the label-permutation ambiguity of audio-only separators. Robustness to degraded attention cues is improved with a mixed curriculum that varies cue quality during training, yielding stable gains even when EEG-speech correlation is moderate. In multi-talker conditions, the model achieves higher signal-to-interference ratio improvements than an audio-only electrodogram baseline while remaining slightly smaller (167k vs. 171k parameters). With 2 ms algorithmic latency and comparable cost, the approach highlights the promise of coupling auditory and neural cues for cognitively adaptive CI processing.

Extremely large antenna arrays envisioned for 6G incurs near-field effect, where steering vector depends on angles and range simultaneously. Polar-domain near-field codebooks can focus energy accurately but incur extra two-dimensional sweeping overhead; compressed-sensing (CS) approaches with Gaussian-masked DFT sensing offer a lower-overhead alternative. This letter revisits near-field beam training using conventional DFT codebooks. Unlike far-field responses that concentrate energy on a few isolated DFT beams, near-field responses produce contiguous, plateau-like energy segments with sharp transitions in the DFT beamspace. Pure LASSO denoising, therefore, tends to over-shrink magnitudes and fragment plateaus. We propose a beam-pattern-preserving beam training scheme for multiple-path scenarios that combines LASSO with a lightweight denoising pipeline: LASSO to suppress small-amplitude noise, followed by total variation (TV) to maintain plateau levels and edge sharpness. The two proximal steps require no near-field codebook design. Simulations with Gaussian pilots show consistent NMSE and cosine-similarity gains over least squares and LASSO at the same pilot budget.

Semantic communication (SemCom) emerges as a transformative paradigm for traffic-intensive visual data transmission, shifting focus from raw data to meaningful content transmission and relieving the increasing pressure on communication resources. However, to achieve SemCom, challenges are faced in accurate semantic quantization for visual data, robust semantic extraction and reconstruction under diverse tasks and goals, transceiver coordination with effective knowledge utilization, and adaptation to unpredictable wireless communication environments. In this paper, we present a systematic review of SemCom for visual data transmission (SemCom-Vision), wherein an interdisciplinary analysis integrating computer vision (CV) and communication engineering is conducted to provide comprehensive guidelines for the machine learning (ML)-empowered SemCom-Vision design. Specifically, this survey first elucidates the basics and key concepts of SemCom. Then, we introduce a novel classification perspective to categorize existing SemCom-Vision approaches as semantic preservation communication (SPC), semantic expansion communication (SEC), and semantic refinement communication (SRC) based on communication goals interpreted through semantic quantization schemes. Moreover, this survey articulates the ML-based encoder-decoder models and training algorithms for each SemCom-Vision category, followed by knowledge structure and utilization strategies. Finally, we discuss potential SemCom-Vision applications.

Generating clinical reports that summarize abnormal patterns, diagnostic findings, and clinical interpretations from long-term EEG recordings remains labor-intensive. We present CELM, the first clinical EEG-to-Language foundation model capable of summarizing long-duration, variable-length EEG recordings and performing end-to-end clinical report generation at multiple scales. CELM integrates pretrained EEG foundation models with language models to enable scalable multimodal learning. We curate a large-scale clinical EEG dataset containing 9,922 reports paired with approximately 11,000 hours of EEG recordings from 9,048 patients to train CELM, and release the benchmark with an automated report-structuring pipeline to facilitate future research. Experimental results show that CELM consistently outperforms existing methods across all evaluation settings. Importantly, we further conduct human evaluation with clinical experts, demonstrating that CELM generates reports that are more clinically coherent, diagnostically reliable, and better aligned with expert interpretation. We release our model and benchmark construction pipeline at https://github.com/Jathurshan0330/CELM.

Compression artifacts from standard video codecs often degrade perceptual quality. We propose a lightweight, semantic-aware pre-processing framework that enhances perceptual fidelity by selectively addressing these distortions. Our method integrates semantic embeddings from a vision-language model into an efficient convolutional architecture, prioritizing the preservation of perceptually significant structures. The model is trained end-to-end with a differentiable codec proxy, enabling it to mitigate artifacts from various standard codecs without modifying the existing video pipeline. During inference, the codec proxy is discarded, and SCENE operates as a standalone pre-processor, enabling real-time performance. Experiments on high-resolution benchmarks show improved performance over baselines in both objective (MS-SSIM) and perceptual (VMAF) metrics, with notable gains in preserving detailed textures within salient regions. Our results show that semantic-guided, codec-aware pre-processing is an effective approach for enhancing compressed video streams.

Neural networks have been successfully applied in various resource-constrained edge devices, where usually central processing units (CPUs) instead of graphics processing units exist due to limited power availability. State-of-the-art research still focuses on efficiently executing enormous numbers of multiply-accumulate (MAC) operations. However, CPUs themselves are not good at executing such mathematical operations on a large scale, since they are more suited to execute control flow logic, i.e., computer algorithms. To enhance the computation efficiency of neural networks on CPUs, in this paper, we propose to convert them into logic flows for execution. Specifically, neural networks are first converted into equivalent decision trees, from which decision paths with constant leaves are then selected and compressed into logic flows. Such logic flows consist of if and else structures and a reduced number of MAC operations. Experimental results demonstrate that the latency can be reduced by up to 14.9 % on a simulated RISC-V CPU without any accuracy degradation. The code is open source at https://github.com/TUDa-HWAI/NN2Logic

High renewable penetration increases the frequency and magnitude of net-load ramps, stressing real-time flexibility. Two commonly deployed remedies are look-ahead economic dispatch (LAED) and ramp products (RPs), yet their operational equivalence under the industry-standard rolling-window dispatch implementation is not well understood. This paper develops linear optimization models for multi-interval LAED and RP-based co-optimization, and proves that an enhanced RP formulation can match LAED's dispatch feasible region at a single time step when additional intertemporal deliverability constraints are enforced. We then show that this equivalence does not generally persist under rolling-window operation because LAED and RP formulations optimize different intertemporal objectives, leading to divergent end-of-window states. Using different test systems under stressed ramping conditions and multiple load levels, we show LAED achieves similar or lower load shedding than RP implementations with the same look-ahead horizon, with the most pronounced differences under high-load, ramp-limited conditions. The study highlights the limitations of current ramp product implementations and suggests enhancements, such as introducing more mid-duration RPs.

Current methods for converting circuit schematic images into machine-readable netlists struggle with component recognition and connectivity inference. In this paper, we present SINA, an open-source, fully automated circuit schematic image-to-netlist generator. SINA integrates deep learning for accurate component detection, Connected-Component Labeling (CCL) for precise connectivity extraction, and Optical Character Recognition (OCR) for component reference designator retrieval, while employing a Vision-Language Model (VLM) for reliable reference designator assignments. In our experiments, SINA achieves 96.47% overall netlist-generation accuracy, which is 2.72x higher than state-of-the-art approaches.

Accurate reconstruction of atmospheric wind fields is essential for applications such as weather forecasting, hazard prediction, and wind energy assessment, yet conventional instruments leave spatio-temporal gaps within the lower atmospheric boundary layer. Unmanned aircraft systems (UAS) provide flexible in situ measurements, but individual platforms sample wind only along their flight trajectories, limiting full wind-field recovery. This study presents a framework for reconstructing four-dimensional atmospheric wind fields using measurements obtained from a coordinated UAS swarm. A synthetic turbulence environment and high-fidelity multirotor simulation are used to generate training and evaluation data. Local wind components are estimated from UAS dynamics using a bidirectional long short-term memory network (Bi-LSTM) and assimilated into a physics-informed neural network (PINN) to reconstruct a continuous wind field in space and time. For local wind estimation, the bidirectional LSTM achieves root-mean-square errors (RMSE) of 0.064 and 0.062 m/s for the north and east components in low-wind conditions, increasing to 0.122 to 0.129 m/s under moderate winds and 0.271 to 0.273 m/s in high-wind conditions, while the vertical component exhibits higher error, with RMSE values of 0.029 to 0.091 m/s. The physics-informed reconstruction recovers the dominant spatial and temporal structure of the wind field up to 1000 m altitude while preserving mean flow direction and vertical shear. Under moderate wind conditions, the reconstructed mean wind field achieves an overall RMSE between 0.118 and 0.154 m/s across evaluated UAS configurations, with the lowest error obtained using a five-UAS swarm. These results demonstrate that coordinated UAS measurements enable accurate and scalable four-dimensional wind-field reconstruction without dedicated wind sensors or fixed infrastructure.

Unmanned aerial vehicles (UAVs) integrated into cellular networks face significant challenges from air-to-ground interference. To address this, we propose a downlink UAV communication system that leverages a fluid antenna system (FAS)- assisted reconfigurable intelligent surface (RIS) to enhance signal quality. By jointly optimizing the FAS port positions and RIS phase shifts, we maximize the achievable rate. The resulting nonconvex optimization problem is solved using successive convex approximation (SCA) based on second-order cone programming (SOCP), which reformulates the constraints into a tractable form. Simulation results show that the proposed algorithm significantly improves both outage probability and achievable rate over conventional fixed-position antenna (FPA) schemes, with particularly large gains in large-scale RIS configurations. Moreover, the algorithm converges rapidly, making it suitable for real-time applications

This paper investigates information freshness in a remote estimation system in which the remote information source is a continuous-time Markov chain (CTMC). For such systems, estimators have been mainly restricted to the class of martingale estimators in which the remote estimate at any time is equal to the value of the most recently received update. This is mainly due to the simplicity and ease of analysis of martingale estimators, which however are far from optimal, especially in query-based (i.e., pull-based) update systems. In such systems, maximum a-posteriori probability (MAP) estimators are optimal. However, MAP estimators can be challenging to analyze in continuous-time settings. In this paper, we introduce a new class of estimators, called structured estimators, which can seamlessly shift from a martingale estimator to a MAP estimator, enabling them to retain useful characteristics of the MAP estimate, while still being analytically tractable. Particularly, we introduce a new estimator termed as the $p$-MAP estimator which is a piecewise-constant approximation of the MAP estimator with finitely many discontinuities, bringing us closer to a full characterization of MAP estimators when modeling information freshness. In fact, we show that for time-reversible CTMCs, the MAP estimator reduces to a $p$-MAP estimator. Using the binary freshness (BF) process for the characterization of information freshness, we derive the freshness expressions and provide optimal state-dependent sampling policies (i.e., querying policies) for maximizing the mean BF (MBF) for pull-based remote estimation of a single CTMC information source, when structured estimators are used. Moreover, we provide optimal query rate allocation policies when a monitor pulls information from multiple heterogeneous CTMCs with a constraint on the overall query rate.

To enhance the efficiency of capacity markets, many electricity markets in the U.S. are adopting or planning to implement marginal capacity accreditation reforms. This paper provides new insights into energy storage capacity accreditation using Marginal Reliability Impact (MRI). We reformulate the commonly used reliability-based storage dispatch model as an optimization problem, enabling direct calculation of the MRI from the Lagrange multipliers, rather than using brute-force perturbation analysis. The analysis demonstrates that the EUE is a piecewise linear function and the storage MRI retains a non-negative property across various system scenarios. We further explore the influence of qualified capacity (QC), storage dispatch rules, and other key factors on storage accreditation, providing practical insights for system operators. Additionally, comparisons of storage capacity accreditation under different reliability criteria offer valuable guidance for policymakers in setting future standards. Numerical results from a modified California system validate our findings and highlight several important phenomena associated with the MRI-based accreditation scheme.

Probabilistic resource adequacy assessment is a cornerstone of modern capacity accreditation. This paper develops a gradient-based framework, in which capacity accreditation is interpreted as the directional derivative of a probabilistic resource adequacy metric with respect to resource capacity, that unifies two widely used accreditation approaches: Effective Load Carrying Capability (ELCC) and Marginal Reliability Impact (MRI). Under mild regularity conditions, we show that marginal ELCC and MRI yield equivalent accreditation factors, while their numerical implementations exhibit markedly different computational characteristics. Building on this framework, we demonstrate how infinitesimal perturbation analysis enables up to a $1000\times$ speedup in gradient estimation for capacity accreditation, and we implement gradient-informed search algorithms that significantly accelerate ELCC computations relative to standard bisection methods. Large-scale Monte Carlo experiments show that MRI achieves substantial runtime reductions compared to ELCC and exhibits greater robustness to perturbation step-size selection. These results provide practical guidance for implementing efficient and scalable capacity accreditation in large-scale power systems.

Voltage (Volt) and reactive-power (VAR) control in transmission networks is critical for reliability and increasingly needs fast, implementable decisions. This paper presents a transmission Volt/VAR Optimization (VVO) framework that co-optimizes discrete control of on-load tap-changing transformers (OLTCs) and capacitor banks (CBs) with AC power flow (ACPF) physics to improve voltage stability and minimize VAR generation. The framework follows a relax-round-resolve pipeline: a continuous relaxation proposes targets, a rounding step selects feasible discrete settings, and a final solve enforces AC power flow physics. Extensive experiments on IEEE, PEGASE, and RTE systems show consistent improvements in voltage and VAR quality metrics with modest generator redispatch while preserving economic operation and achieving compatible runtimes with real-time transmission operations.

Feature coding for machines (FCM) is a lossy compression paradigm for split-inference. The transmitter encodes the outputs of the first part of a neural network before sending them to the receiver for completing the inference. Practical FCM methods “sandwich” a traditional codec between pre- and post-processing neural networks, called wrappers, to make features easier to compress using video codecs. Since traditional codecs are non-differentiable, the wrappers are trained using a proxy codec, which is later replaced by a standard codec after training. These codecs perform rate-distortion optimization (RDO) based on the sum of squared errors (SSE). Because the RDO does not consider the post-processing wrapper, the inner codec can invest bits in preserving information that the post-processing later discards. In this paper, we modify the bit-allocation in the inner codec via a wrapper-aware weighted SSE metric. To make wrapper-aware RDO (WA-RDO) practical for FCM, we propose: 1) temporal reuse of weights across a group of pictures and 2) fixed, architecture- and task-dependent weights trained offline. Under MPEG test conditions, our methods implemented on HEVC match the VVC-based FCM state-of-the-art, effectively bridging a codec generation gap with minimal runtime overhead relative to SSE-RDO HEVC.

Urban mobility systems are transitioning toward electric, on-demand services, creating operational challenges for fleet management under energy and service-quality constraints. The Electric Dial-a-Ride Problem (E-DARP) extends the classical dial-a-ride problem by incorporating limited battery capacity and nonlinear charging dynamics, increasing computational complexity and limiting the scalability of exact methods for real-time use. This paper proposes a deep reinforcement learning approach based on an edge-centric graph neural network encoder and an attention-driven route construction policy. By operating directly on edge attributes such as travel time and energy consumption, the method captures non-Euclidean, asymmetric, and energy-dependent routing costs in real road networks. The learned policy jointly optimizes routing, charging, and service quality without relying on Euclidean assumptions or handcrafted heuristics. The approach is evaluated on two case studies using ride-sharing data from San Francisco. On benchmark instances, the method achieves solutions within 0.4% of best-known results while reducing computation times by orders of magnitude. A second case study considers large-scale instances with up to 250 request pairs, realistic energy models, and nonlinear charging. On these instances, the learned policy outperforms Adaptive Large Neighborhood Search (ALNS) by 9.5% in solution quality while achieving 100% service completion, with inference times under 10 seconds compared to hours for the metaheuristic. Finally, sensitivity analyses quantify the impact of battery capacity, fleet size, ride-sharing capacity, and reward weights, while robustness experiments show that deterministically trained policies generalize effectively under stochastic conditions.

Movable antennas (MA) have gained significant attention in recent years to overcome the limitations of extremely large antenna arrays in terms of cost and power consumption. In this paper, we investigate the use of MA arrays at the base station (BS) for angle-of-departure (AoD) estimation under uncertainty in the user equipment (UE) location. Specifically, we (i) derive the theoretical performance limits through the Cramér-Rao bound (CRB) and (ii) optimize the antenna positions to ensure robust performance within the UE's uncertainty region. Numerical results show that dynamically optimizing antenna placement by explicitly considering the uncertainty region yields superior performance compared to fixed arrays, demonstrating the ability of MA systems to adapt and outperform conventional arrays.

An agent operating in an unknown dynamical system must learn its dynamics from observations. Active information gathering accelerates this learning, but existing methods derive bespoke costs for specific modeling choices: dynamics models, belief update procedures, observation models, and planners. We present a unifying framework that decouples these choices from the information-gathering cost by explicitly exposing the causal dependencies between parameters, beliefs, and controls. Using this framework, we derive a general information-gathering cost based on Massey's directed information that assumes only Markov dynamics with additive noise and is otherwise agnostic to modeling choices. We prove that the mutual information cost used in existing literature is a special case of our cost. Then, we leverage our framework to establish an explicit connection between the mutual information cost and information gain in linearized Bayesian estimation, thereby providing theoretical justification for mutual information-based active learning approaches. Finally, we illustrate the practical utility of our framework through experiments spanning linear, nonlinear, and multi-agent systems.

The performance of speaker verification systems degrades significantly under language mismatch, a critical challenge exacerbated by the field's reliance on English-centric data. To address this, we propose the TidyVoice Challenge for cross-lingual speaker verification. The challenge leverages the TidyVoiceX dataset from the novel TidyVoice benchmark, a large-scale, multilingual corpus derived from Mozilla Common Voice, and specifically curated to isolate the effect of language switching across approximately 40 languages. Participants will be tasked with building systems robust to this mismatch, with performance primarily evaluated using the Equal Error Rate on cross-language trials. By providing standardized data, open-source baselines, and a rigorous evaluation protocol, this challenge aims to drive research towards fairer, more inclusive, and language-independent speaker recognition technologies, directly aligning with the Interspeech 2026 theme, "Speaking Together."

Diffusion speech enhancement on discrete audio codec features gain immense attention due to their improved speech component reconstruction capability. However, they usually suffer from high inference computational complexity due to multiple reverse process iterations. Furthermore, they generally achieve promising results on non-intrusive metrics but show poor performance on intrusive metrics, as they may struggle in reconstructing the correct phones. In this paper, we propose DisContSE, an efficient diffusion-based speech enhancement model on joint discrete codec tokens and continuous embeddings. Our contributions are three-fold. First, we formulate both a discrete and a continuous enhancement module operating on discrete audio codec tokens and continuous embeddings, respectively, to achieve improved fidelity and intelligibility simultaneously. Second, a semantic enhancement module is further adopted to achieve optimal phonetic accuracy. Third, we achieve a single-step efficient reverse process in inference with a novel quantization error mask initialization strategy, which, according to our knowledge, is the first successful single-step diffusion speech enhancement based on an audio codec. Trained and evaluated on URGENT 2024 Speech Enhancement Challenge data splits, the proposed DisContSE excels top-reported time- and frequency-domain diffusion baseline methods in PESQ, POLQA, UTMOS, and in a subjective ITU-T P.808 listening test, clearly achieving an overall top rank.

Laser inter-satellite links (LISLs) of low Earth orbit (LEO) mega-constellations enable high-capacity backbone connectivity in non-terrestrial networks, but their management is challenged by limited laser communication terminals, mechanical pointing constraints, and rapidly time-varying network topologies. This paper studies the joint problem of LISL connection establishment, traffic routing, and flow-rate allocation under heterogeneous global traffic demand and gateway availability. We formulate the problem as a mixed-integer optimization over large-scale, time-varying constellation graphs and develop a Lagrangian dual decomposition that interprets per-link dual variables as congestion prices coordinating connectivity and routing decisions. To overcome the prohibitive latency of iterative dual updates, we propose DeepLaDu, a Lagrangian duality-guided deep learning framework that trains a graph neural network (GNN) to directly infer per-link (edge-level) congestion prices from the constellation state in a single forward pass. We enable scalable and stable training using a subgradient-based edge-level loss in DeepLaDu. We analyze the convergence and computational complexity of the proposed approach and evaluate it using realistic Starlink-like constellations with optical and traffic constraints. Simulation results show that DeepLaDu achieves up to 20% higher network throughput than non-joint or heuristic baselines, while matching the performance of iterative dual optimization with orders-of-magnitude lower computation time, suitable for real-time operation in dynamic LEO networks.

Low Earth orbit (LEO) mega-constellations greatly extend the coverage and resilience of future wireless systems. Within the mega-constellations, laser inter-satellite links (LISLs) enable high-capacity, long-range connectivity. Existing LISL schemes often overlook mechanical limitations of laser communication terminals (LCTs) and non-uniform global traffic profiles caused by uneven user and gateway distributions, leading to suboptimal throughput and underused LCTs/LISLs – especially when each satellite carries only a few LCTs. This paper investigates the joint optimization of LCT connections and traffic routing to maximize the constellation throughput, considering the realistic LCT mechanics and a snapshot-level spatial traffic profile. We show that the resulting per-snapshot formulation is a mixed-integer program coupling LCT connections with flow-rate variables under link capacity constraints, and the associated fixed-snapshot abstract decision problem is NP-hard. Due to its intractability, we resort to relaxing the coupling constraints via Lagrangian duality, decomposing the problem into a weighted graph-matching for LCT connections, weighted shortest-path routing tasks, and a linear program for rate allocation. Here, Lagrange multipliers reflect congestion weights between satellites, jointly guiding the matching, routing, and rate allocation. Subgradient descent optimizes the multipliers, with proven convergence. Simulations using real-world constellation and terrestrial data show that our methods substantially improve network throughput by up to $35\%$–$145\%$ over existing non-joint approaches.

Massive multiple-input multiple-output (mMIMO) downlink precoding offers high spectral efficiency but remains challenging to deploy in practice because near-optimal algorithms such as the weighted minimum mean squared error (WMMSE) are computationally expensive, and sensitive to SNR and channel-estimation quality, while existing deep learning (DL)-based solutions often lack robustness and require retraining for each deployment site. This paper proposes a plug-and-play precoder (PaPP), a DL framework with a backbone that can be trained for either fully digital (FDP) or hybrid beamforming (HBF) precoding and reused across sites, transmit-power levels, and with varying amounts of channel estimation error, avoiding the need to train a new model from scratch at each deployment. PaPP combines a high-capacity teacher and a compact student with a self-supervised loss that balances teacher imitation and normalized sum-rate, trained using meta-learning domain-generalization and transmit-power-aware input normalization. Numerical results on ray-tracing data from three unseen sites show that the PaPP FDP and HBF models both outperform conventional and deep learning baselines, after fine-tuning with a small set of local unlabeled samples. Across both architectures, PaPP achieves more than 21$\times$ reduction in modeled computation energy and maintains good performance under channel-estimation errors, making it a practical solution for energy-efficient mMIMO precoding.

We design a variational state estimation (VSE) method that provides a closed-form Gaussian posterior of an underlying complex dynamical process from (noisy) nonlinear measurements. The complex process is model-free. That is, we do not have a suitable physics-based model characterizing the temporal evolution of the process state. The closed-form Gaussian posterior is provided by a recurrent neural network (RNN). The use of RNN is computationally simple in the inference phase. For learning the RNN, an additional RNN is used in the learning phase. Both RNNs help each other learn better based on variational inference principles. The VSE is demonstrated for a tracking application - state estimation of a stochastic Lorenz system (a benchmark process) using a 2-D camera measurement model. The VSE is shown to be competitive against a particle filter that knows the Lorenz system model and a recently proposed data-driven state estimation method that does not know the Lorenz system model.

A large number of works view the automatic assessment of speech from an utterance- or system-level perspective. While such approaches are good in judging overall quality, they cannot adequately explain why a certain score was assigned to an utterance. frame-level scores can provide better interpretability, but models predicting them are harder to tune and regularize since no strong targets are available during training. In this work, we show that utterance-level speech quality predictors can be regularized with a segment-based consistency constraint which notably reduces frame-level stochasticity. We then demonstrate two applications involving frame-level scores: The partial spoof scenario and the detection of synthesis artefacts in two state-of-the-art text-to-speech systems. For the latter, we perform listening tests and confirm that listeners rate segments to be of poor quality more often in the set defined by low frame-level scores than in a random control set.

This paper addresses multi-objective conflicts and catastrophic forgetting in uncrewed aerial vehicle (UAV) networks across dynamic spatiotemporal environments. Conventional multi-agent reinforcement learning (MARL) algorithms suffer from severe policy degradation during sequential task transitions. We propose a spatiotemporal continual federated learning (SCFL) framework driven by the group-decoupled multi-agent proximal policy optimization (G-MAPPO) algorithm. SCFL incorporates a three-stage geometric alignment mechanism: it resolves local gradient conflicts via group-decoupled policy optimization (GDPO), mitigates spatial Non-Independent and Identically Distributed (non-IID) client drift through adaptive cosine aggregation, and suppresses inter-task interference via global temporal orthogonal projection without raw experience replay. Evaluations show that SCFL achieves superior robustness over federated baselines, maintaining spatial service reliability above 0.95 and a load balancing index of approximately 0.95 during non-stationary transitions. A longitudinal self-degradation analysis further shows that SCFL preserves historical knowledge with near-zero performance variation in spatial reliability and QoS under moderate loads from 40 to 120 users, while revealing its operating boundary under extreme congestion with 140 users due to hard projection constraints. The framework provides a scalable, communication-efficient approach for autonomous aerial network orchestration.