Papers

Humans rely on gaze, head movements, and visual cues to attend to speakers in noisy environments, yet auditory attention decoding (AAD) has been studied primarily using electroencephalography (EEG). We introduce the Multimodal Auditory-attention Egocentric Speech-TRacking Open (MAESTRO) corpus, the first AAD dataset to simultaneously record EEG, eye gaze, pupillometry, egocentric video, and head inertial measurement unit (IMU) data. MAESTRO includes four competing speakers and background noise across multiple signal-to-noise ratio (SNR) conditions, enabling attention decoding under realistic listening scenarios. Through a four-speaker attention decoding benchmark, we show that combining behavioral and physiological signals improves decoding performance over EEG-only approaches, enabling future advances in multimodal auditory attention decoding. These findings open the door to new applications, analyses, and methodological advances in multimodal AAD. The complete dataset is publicly available at https://huggingface.co/datasets/aspire-osu/maestro-eeg-dataset . The official code repository is available at https://github.com/ASPIRE-OSU/MAESTRO .

While modern text-to-speech (TTS) systems generate highly natural speech and support inline non-verbal vocalization (NVV) tags, accurate control over these events remains challenging. A key gap is the lack of established post-training methods for non-verbal control in continuous autoregressive flow-matching TTS. To this end, we present NVAlign, a direct-gradient post-training framework for NVV tag-following in this architecture. We first perform supervised fine-tuning (SFT) of TTS models and an NVV-aware automatic speech recognition (NV-ASR) model on NVV-annotated speech, then freeze the NV-ASR model to serve as the reward model for post-training. A two-step gradient surrogate enables efficient reward backpropagation through the flow-matching sampler to jointly update the autoregressive backbone and acoustic flow head. Fidelity penalties and reference-velocity regularization help preserve speaker similarity and speech quality. Results from NVV-SuperBench and human listening evaluations show that NVAlign improves tag-following accuracy over SFT and Flow-GRPO baselines. These findings demonstrate that direct reward-gradient optimization can improve non-verbal control in continuous autoregressive flow-matching TTS. Audio samples are available at https://nvalign.github.io/.

Open-vocabulary keyword spotting (KWS) must detect arbitrary keywords without retraining. Cross-attention-based models achieve state-of-the-art performance but require pairwise interaction between the audio and each keyword, recomputing the fused representation for every audio–keyword pair. Embedding-based models avoid this computational cost through independent encoding and similarity scoring, yet remain inferior on challenging benchmarks such as LibriPhrase-hard. We show that this gap is primarily a training issue rather than a limitation of model capacity. While contrastive learning encourages correct keywords to rank above negatives, it does not explicitly calibrate absolute similarity scores, which are critical for applying a fixed threshold across keyword detection. We propose Calibrated, Order-aware Detection KWS (CORD-KWS), a training framework that introduces two complementary objectives while retaining the same encoders, cosine scoring, and O(1) keyword enrollment: a calibrated detection head that supervises absolute scores, and a frame-level CTC objective that preserves phonetic order information. CORD-KWS achieves EERs of 0.43% and 9.64% on LibriPhrase-easy and LibriPhrase-hard, respectively, outperforming existing embedding- and cross-attention-based models.

Rational stability is a quantitative stability notion for nonlinear ordinary differential equations (ODEs), under which convergence is characterized by a rationally-in-time decaying bound on solutions. Although Lyapunov characterizations of rational stability have been proposed, no existing condition has been shown to exactly characterize rational rate performance. The paper resolves this issue by providing an exact converse Lyapunov characterization of regional rational stability with prescribed rate performance. To test the resulting converse Lyapunov conditions numerically, a tiered set of convex relaxations is proposed which can be enforced using Sum-of-Squares (SOS) programming. A conjectured conservatism scaling factor in rate performance is associated with each tier and validated numerically. The first proposed tier appears to be novel and is shown to outperform classical SOS-based approaches. Numerical examples are used to validate the results and compute nested regions of state-space on which prescribed levels of performance can be guaranteed.

Dormant Awakening (DA) specializes a magnetic resonance imaging (MRI) reconstruction model by fitting a small set of zero weights to one labeled slice. Across fourteen U-shaped convolutional network (U-Net) and vision transformer (ViT) sources, fitting 1,004 or 10,560 weights gives mean peak signal-to-noise ratio (PSNR) gains of .430 and .509 dB. We analyze how scaling the fitted output correction changes evaluation PSNR, then use the calibration correction size to cap changes on new images. The cap bounds changes in root mean squared error (RMSE) relative to the source without an evaluation reference. At 10% adaptation budget, we compare fixed halving and dynamic capping for DA, unrestricted sparse adaptation and low-rank adaptation (LoRA). The controls increase both measured calibration-evaluation correlations across all six tested architecture-adapter settings. Experiments concern one constructed fastMRI-to-M4Raw shift with fixed source checkpoints and participant panels.

Pruning reduces the number of weights in magnetic resonance imaging (MRI) reconstruction networks. Equal sparsity, however, can retain weights with different computational roles and different compatibility with the trained network. We study these effects across 120 convolutional U-Net and vision transformer models using controlled edits evaluated before retraining. In U-Net, preserving high-resolution computation improves reconstruction at equal deletion counts across all tested sparsity levels. Equal-operation controls reveal additional sensitivity of the first convolution, which operation count alone cannot explain. In transformers, retained pretrained weight energy correlates with quality within sparsity levels, yet the same masks reverse their reconstruction ranking when the surrounding trained weights change. At 90% sparsity, matching intermediate output scales largely removes a high-energy-mask penalty while retaining an advantage for the network's original mask. Thus sparsity alone does not characterize reconstruction quality: architecture-specific descriptors are informative, but mask quality can still depend on the surrounding trained network.

In this work, we explore MammoClaw, a training-free agent framework that leverages frozen MLLMs for mammography analysis. To support agentic investigation, we equip the agent with lightweight mammography-specific tools for targeted image analysis, including ROI, paired-view, and contralateral-breast examination. MammoClaw iteratively gathers evidence through these tools, while skill evolution enables non-parametric adaptation by transforming failed trajectories into reusable guidance for later runs. We evaluate the framework on BI-RADS assessment and breast density estimation tasks. In our experiments, we find that tools alone do not reliably improve performance, whereas evolved skills can improve tool-use behavior and performance in some settings. Beyond these results, MammoClaw enables transparent inspection of evidence acquisition, tool interactions, and failure modes, facilitating the analysis and auditing of agent behavior. We view this work as an exploratory study of training-free, self-evolving agentic approaches for mammography and hope it provides a concrete starting point for future work on mammography-specific tools and self-evolution mechanisms. We release our code at https://krishnakanthnakka.github.io/mammoclaw.

Stacked intelligent metasurfaces process the transmitted field layer by layer, and fluid antennas make the position of every radiator a design variable. Combined, they pack radiators at sub-wavelength spacings within and across layers, where mutual coupling governs the physics that current models omit. This paper develops a coupling-consistent model of a fluid transmit layer illuminating a stack of passive fluid beyond-diagonal layers. The multiport impedance matrix of any port constellation is shown to be the impedance kernel sampled at the three-dimensional separations, in one closed spherical-Hankel form whose in-plane restriction is intra-layer coupling and whose axial restriction replaces the scalar Rayleigh–Sommerfeld propagator. Only the transmitter is driven: the layer currents are induced, and the familiar cascade is the single-pass limit of one matrix inversion. In the wavenumber domain the light circle still separates radiation from reaction, now with a plane-wave propagator attached, every Fourier mode sees a transmission line loaded by the layers, and an efficiency identity shows that what the stack costs in efficiency is set by its total induced-current norm, not by its layer count. Because the load is never assumed layer-diagonal, interconnections may join atoms on different sheets, and vertical networks repeated at every site are spectrally local. Alternating optimization designs precoders, positions, and loads with closed-form gradients. Numerically, fluid ports on a $\lambda/8$ grid perform within a few per cent of continuously movable ones, while a design based on the conventional cascade model can do worse than open-circuiting the stack.

Many voice interaction applications require exact control over both the content and delivery of responses, typically using pre-recorded lines. Recent full-duplex models respond with low latency but cannot guarantee exact content or reproduce a specific recorded performance, while cascaded systems can be constrained to predefined responses at the cost of additional latency. We propose RePlay, a spoken dialogue system adapted from PersonaPlex that handles multi-turn conversations by retrieving and playing pre-recorded lines. Using probing, we identify the layer and frame at which the upcoming response becomes recoverable, and use this hidden state as the retrieval query. RePlay retains only the layers up to that point and replaces text and speech generation with lightweight turn-taking and retrieval heads. In simulated multi-turn interviews, RePlay reaches a median latency of 383 ms, 3 to 7 times lower than ASR-LLM cascades of comparable dialogue quality, at the cost of lower exact-line accuracy. In a user study, participants preferred RePlay in 63% of ratings versus 12% for a fast cascade with a small LLM (p = 0.008), and showed a non-significant preference (46% vs. 21%) over a slower cascade with a stronger LLM.

Online learning in non-stationary environments requires models to adapt rapidly from streaming data under strict computational constraints. A principled approach casts online learning as Bayesian state tracking, where model parameters are updated sequentially via Bayesian filtering. However, applying Bayesian filters directly to modern deep models is computationally prohibitive due to the high dimensionality of parameter space, forcing existing methods to rely on restrictive approximations or manually designed low-dimensional subspaces. In this work, we identify the absence of a suitable low-dimensional dynamical representation as the core bottleneck in Bayesian filtering-based online learning. Accordingly, we propose Adaptive Update through Representation Adaptation (AURA), a meta-learning framework that learns offline a low-dimensional latent state-space model governing the evolution of optimal model parameters under distribution shift. Online adaptation is then performed via extended Kalman filtering in this learned latent space followed by reconstruction of the full model parameters through a learned lifting map, enabling efficient single-step online adaptation while preserving model expressiveness. Evaluated on online adaptation of neural wireless receivers under time-varying channels and on non-stationary image classification, AURA shows substantial improvements in adaptation speed, accuracy, and computational efficiency over existing online learning and Bayesian filtering baselines, demonstrating that an adaptation-aware latent geometry is beneficial for effective Bayesian online learning in high-dimensional models.

Text-to-audio (TTA) generation has advanced rapidly in generation quality and instruction following. However, representative systems often require around a billion parameters, limiting deployment on resource-constrained devices. This paper introduces TinyAudio, a compact flow-matching-based TTA model for low-resource deployment. At its core, TinyAudio uses TA-DiT, a 35M single-stream flow-matching Transformer. TinyAudio also includes TA-CLAP, a 32M audio-aligned text encoder, and TA-VAE, whose 20M decoder reconstructs 44.1 kHz audio from compressed latents. TinyAudio has only 87M parameters in total, over 90% fewer than representative billion-parameter pipelines, and uses 0.48 GB peak GPU memory. TinyAudio achieves competitive generation quality on AudioCaps and TTA-Bench. We further introduce TinyAudio-MF, a MeanFlow-accelerated model that enables real-time generation with a four-core CPU quota. Our results demonstrate a practical quality-footprint trade-off for low-resource TTA deployment.

This study evaluates a mathematical model of syllable production using an EMA dataset from a recently published study, which proved highly compatible with the model's architecture. The data set consists of regularly structured French phrases of equal duration and reduced phonetic and syllabic complexity. A dedicated procedure was used to transform EMA recordings into Maeda parameters. The Model-generated trajectories were then realigned with these transformed data using Canonical Time Warping (CTW). Statistical validation was conducted via a permutation test, comparing alignment quality against a surrogate condition with phonetically incongruent model outputs. The results show significantly better alignment for phonetically congruent pairings, revealing a strong structural correspondence between the model and the articulatory data. These findings provide experimental support for the representational validity of the model.

The goal of personalized immunogenomics is to recover an individual's germline immunoglobulin gene segments from 'repertoire sequences' altered by trimming, extension, and mutation. In this work, we study the fundamental trace complexity (or the number of samples/traces required for accurate reconstruction) of D-gene reconstruction under three biologically motivated trace-generation models introduced by Bhardwaj et al. (2021) and develop practical algorithms for reconstruction problems left open in the original work. First, for the TrimSuffixAndExtend model, we establish the optimal trace complexity to be {\Theta(n)}, and develop a low-complexity Prefix-Filtered Mode (PFM) decoder that achieves this scaling. Second, for the closely related two-sided TrimAndExtend model, we show that the optimal trace complexity is instead {\Theta(n^2)}, and is achieved by the simple Bit-Wise Mode (BWM) decoder. Third, for the SuffixExtend-t(TrimSuffix) model, we establish polynomially separated lower and upper bounds on trace complexity. Our results follow from information-theoretic lower bounds coupled with tight analyses of the proposed reconstruction algorithms.

In daily life, people hear speech, footsteps, and music around them. We can often recognize these sounds and judge where they come from. Each sound source can be shown on a separate acoustic map, a rectangular image covering $360^{\circ}$ horizontally and $180^{\circ}$ vertically. The map shows the directions occupied by the source as a region and the sound energy within that region. A class label identifies the sound. Predicting these labeled acoustic maps from audio is called semantic acoustic imaging. Such maps could help robots perceive their surroundings and allow augmented reality displays to show sound regions and classes over the real world. Existing models can recognize sound classes and estimate a direction for each source. However, a direction alone does not describe the source region or its energy. Acoustic imaging must also distinguish sound sources in nearby directions, while the number of active sources and the regions they occupy can change over time. We therefore propose the Semantic Acoustic Imaging Detector (SAID), which predicts a separate labeled acoustic map for each active source from audio. First, we pretrain Audio2Sph, SAID's audio encoder, through sound energy estimation across directions without class labels. Then, we train the complete SAID model to predict source regions, energy, and classes together. We also develop a pipeline that generates simulated recordings for pretraining and supports fine-tuning on real recordings. On the official DCASE2026 Task 3 Track A evaluation set, our submitted system ranks first with 0.1080 macro-averaged mean average precision (Macro mAP) and 0.3962 Macro Pearson $r$. Demos and code are provided at https://github.com/IN03X/SAID.

Task-agnostic assistive exoskeleton control based on human intention offers greater flexibility than conventional approaches that rely on predefined tasks or motion patterns. Human joint torque estimation enables task-agnostic assistance by characterizing user actions. Physics-consistent methods such as Deep Lagrangian Networks (DeLaN) have been applied to estimate the human torques in multi-user settings, but existing approaches cannot adapt to a specific user without retraining, and do not account for intermittent contacts during locomotion. We propose ExoLaN, a Context-Aware DeLaN for human-exoskeleton interaction that learns the full coupled system dynamics while adapting to changes in interaction context. ExoLaN combines temporal context with partial contact-force measurements from force-sensitive insoles to infer latent dynamics embeddings and estimate generalized contact torques. On seven unseen users performing 21 unseen tasks, ExoLaN reduces torque estimation MSE by 7% compared to a black-box baseline. Beyond inverse dynamics, ExoLaN serves as a unified model that also enables accurate forward prediction: training with a multi-step prediction loss reduces acceleration MSE by 59% and long-horizon position and velocity errors by 60% and 93%, respectively, compared with a single-step loss. Moreover, the learned latent context captures task information without explicit task labels, making it a promising signal for task-aware assistive control.

Multimodal inference in safety-critical applications, such as autonomous driving and robot navigation, requires heterogeneous sensor observations to reach an edge server within a task-prescribed temporal window. Referenced to the event or state being inferred, this window ends at the latest time an observation remains useful for the current decision. Since sensor readiness times, data volumes, wireless channel conditions, and task relevance vary across sensors, maximising network throughput does not necessarily minimise the fused prediction error when the window closes. We cast multimodal uplink scheduling as sequential wireless evidence acquisition and, for a linear minimum mean-square error (LMMSE) fusion model, derive a conditional evidence gain metric from the reduction in residual-error volume. Defined through second-order statistics, this metric applies beyond Gaussian models and coincides with conditional mutual information when the target and prediction errors are jointly Gaussian. We then develop MIRA, a greedy task-driven maximum information-rate allocation policy that combines conditional evidence gain with each sensor's channel state information and remaining data volume, while updating sensor relevance as evidence is acquired. Experiments on synthetic classification, human activity recognition, and vehicle-trajectory regression show that MIRA outperforms both relevance-only and channel-only scheduling. Relative to the former, MIRA improves classification accuracy by up to 70% and reduces regression mean-square error (MSE) by up to 5.5%. Relative to the latter, it requires 58% less acquisition time to attain an 80% target accuracy, while achieving up to 95% higher accuracy and 9% lower regression MSE. These gains are achieved without maximising received-data volume.

This paper presents mathematically principled filters for multiple-model systems with states of different dimensionality, specifically the variable-dimension interacting multiple model (VD-IMM) filter and the variable-dimension generalised pseudo-Bayesian filter of order 2 (VD-GPB2). To do so, we first provide a Bayesian modelling of a variable dimensional dynamic system, and its measurements. Then, for variable dimensional linear Gaussian dynamic and measurement models, the VD-IMM filter is derived by assuming a posterior that has a Gaussian density for each mode, and then performing a Kullback-Leibler Divergence (KLD) minimisation after each prediction step to keep the Gaussian density form for each mode. Subsequently, the Bayesian update step is performed, keeping a Gaussian density for each mode. The VD-GPB2 filter considers the same model as the VD-IMM filter and also assumes that the posterior for each mode is Gaussian. In contrast, the VD-GPB2 filter propagates a Gaussian mixture for each mode in the prediction step. In the update step, the VD-GPB2 filter performs a KLD minimisation to have a Gaussian density for each mode. Simulation results show the benefits of the VD-IMM and VD-GPB2 filters compared to previous alternatives.

Auditory Attention Detection (AAD) utilizes electroencephalographic (EEG) signals to identify a target speaker in a multi-speaker environment. Despite considerable progress, existing deep learning architectures often lack explicit mechanisms for handling noisy EEG data. To address this limitation, we propose Auditory Focus Attention Networks (AFA-Net), a machine learning framework that replaces vanilla attention with a simple yet flexible differential attention mechanism to help focus on task-relevant neural activity. AFA-Net achieves an upward accuracy of 96.8% at the 2s decision window, while using substantially fewer parameters than most existing methods. To the best of our knowledge, AFA-Net is among the first frameworks to explicitly try to combat EEG noise to improve AAD.

Multimodal emotion recognition (MER) increasingly pairs EEG with speech, treating internal neural signals and external vocal expression as informative views of affect. In practice, naive fusion underperforms the stronger single modality, because EEG artifacts inject noise that corrupts the shared representation. We introduce EmoSpeechBrain, a multimodal framework built on the insight that noise suppression is a precondition for effective fusion. Its EEG encoder uses differential attention, taking the difference between two attention maps to cancel shared noise and isolate discriminative neural activity. An attention-based gating adapter aligns both modalities in a shared space and weights each one's contribution to the prediction. On two datasets - PME4 and EAV, EmoSpeechBrain improves MER accuracy by up to 12.9% over other state-of-the-art (SOTA) EEG encoders, and surpasses unimodal speech and EEG baselines by up to 13.1% and 23.1%. These results show that once EEG noise is suppressed, fusion delivers gains that naive combination cannot.

Congenital heart disease (CHD) is the most common birth defect, yet a large fraction of cases remain undetected on prenatal ultrasound, in part because current artificial-intelligence methods assume that the key diagnostic frames have already been isolated from a study, by a clinician or by a view classifier. We remove that assumption and address CHD screening directly at the level of the whole ultrasound study. We propose a two-stage framework that first learns transferable frame representations by self-supervised masked-autoencoder pre-training on unlabeled fetal ultrasound, then identifies cardiac frames with a disease-robust module and aggregates them with a transformer-based multiple instance learning (MIL) model that produces a case-level diagnosis from study-level labels alone. The model further returns its highest-scoring frames for clinician review, and a hierarchical head separates critical from non-critical CHD. On the internal test set of our multi-source development cohort (FUSE), the proposed cardiac-gated MIL model reaches an area under the curve (AUC) of 0.985 with a specificity of 0.990, outperforming the reproduced NATMED ensemble (AUC 0.861, specificity 0.600) and the FetalCLIP foundation model (AUC 0.867, specificity 0.710). On an independent external cohort, all models initially perform near chance, but label-free CORAL adaptation raises the proposed model from an AUC of 0.513 to 0.944, whereas whole-study and view-dependent baselines do not recover. These results indicate that whole-study MIL with disease-robust cardiac-frame identification is an accurate and deployable route to prenatal CHD screening.

Tree-Adjoining Grammars (TAGs) have recently been introduced to Nonlinear System Identification (NLSI) as a means of encoding an entire model class as a finite set of grammatical rules, from which candidate models are assembled as trees. Existing TAG-based identifiers rely on evolutionary optimisation and return point estimates of the model structure. This paper instead proposes the TAG framework within a Bayesian setting. A generative prior is defined over tree structures and their parameters, and a Reversible-Jump MCMC sampler with structure-preserving tree moves is used to infer the joint posterior over model structure, parameters and predictions. Two training objectives are considered; that is, a one-step-ahead objective with conjugate parameter proposals, and a simulation-based objective handled by likelihood-free inference. The approach is validated on a simulated polynomial NARX system, the Silverbox benchmark, and wave-loading data from the Christchurch Bay Tower, where embedding Morison's equation as a fixed initial tree yields a grey-box model that outperforms the physics-driven baseline. The results demonstrate that Bayesian TAGs are well suited to quantifying uncertainty in equation discovery for dynamical systems and to fitting physics-informed models.

High-frequency industrial cyber-physical systems (CPS) stream continuous sensor telemetry at sub-10-second intervals. Multi-horizon forecasting over hours is required for predictive maintenance and operational control. However, deploying machine learning models over extended horizons (H = 2,160 steps at dt = 5 s) encounters the Dual-Timescale Forecasting Dilemma: the trade-off between short-horizon kinetic momentum and long-horizon diurnal thermodynamic equilibrium. Autoregressive rollouts suffer from compounding error drift, while direct multi-output projectors exhibit variance explosion and noise extrapolation. In this work, we prove that autoregressive mutual information decays exponentially toward zero over extended lead times, and that the optimal minimum-variance estimator converges asymptotically to the periodic diurnal baseline (Lemma 1 and Theorem 1). Guided by these bounds, we propose the Dynamic Asymptotic Decomposition (DAD) framework. The architecture couples a continuous sigmoidal authority transition schedule \(\alpha(h)\) with monotonic L2 regularization scaling \(\lambda(h)\), transferring prediction authority from kinetic autoregression to diurnal equilibrium while enforcing physical non-negativity. Evaluated across a 30-day industrial telemetry stream (N = 518,400 steps) over 16 walk-forward test windows, DAD achieves an overall Mean Absolute Error (MAE) of 0.1764, representing a 78.3% error reduction over Deep LSTM (0.8143), 60.2% over PatchTST (0.4431), and 20.8% over DLinear (0.2227). Computational profiling demonstrates edge execution in 0.72 ms with 0.203 MFLOPs and 72k parameters on a single CPU core (over 3400x faster than Transformer baselines), validated by an automated dual-stage CI/CD quality assurance harness and physical fail-safe architecture.

Speech-based screening is a promising, non-invasive approach for detecting Alzheimer's disease and related cognitive risks. However, models trained on a single domain often generalize poorly to unseen languages, tasks, or recording protocols. This paper investigates this deployment gap using a leave-one-corpus-out evaluation across four distinct datasets. Among 70 interpretable speech and language features, 59 exhibit direction conflicts between healthy control and cognitive risk groups across corpora, with pause, silence, and speech rate showing high protocol sensitivity. Furthermore, while the XLM-R text baseline achieves strong average performance, its Area Under the ROC Curve (AUC) drops to 0.520 on the weakest held-out domain. A standard GroupDRO baseline reaches a 0.766 mean speaker AUC and a 0.504 worst-domain AUC under the same protocol. To address this, we propose a fusion method that integrates XLM-R text baseline scores with evidence anchors selected during training. Balanced fusion achieves a 0.785 mean speaker AUC, while anchor-heavy fusion raises the worst-case speaker AUC to 0.615. This work highlights the need to audit feature transferability and report worst-case domain robustness in cognitive speech screening.

Digital twins are increasingly used in diabetes research, but reproducing individual glucose dynamics requires accurate identification of glucoregulatory model parameters. Traditional sensitivity analysis can identify influential parameters, yet a ranking based on limited conditions may miss parameters that matter during specific disturbances or for particular individuals. We therefore examine both the magnitude and timing of parameter influence across dynamic input-output conditions and assess whether a common ranking holds across participants. We analyze the Hovorka glucoregulatory model using data from 192 participants receiving automated insulin delivery therapy in the Type 1 Diabetes and Exercise Initiative dataset. We extend Sobol sensitivity analysis to time series and rank parameter influence under four conditions: full-day profiles, isolated meal disturbances, insulin bolus injections, and postprandial responses. We combine the condition-specific results into a global ranking and use it to select parameters for participant-specific identification. Compared with population parameters, identification restricted to the sensitivity-derived subset reduces the root mean square error of 60-minute glucose predictions by 60%, to approximately 31 mg/dL. These findings suggest that a global ranking can capture parameter influence across individuals and dynamic conditions. By narrowing the parameters requiring identification, this approach reduces computational cost and could accelerate the development of personalized diabetes digital twins.

Reliable high-dimensional variable selection requires scalable error-controlling methods. The Terminating-Random Experiments (T-Rex) selector estimates the false discovery rate (FDR) by aggregating early-terminated forward-selection paths in which predictors compete with synthetic dummies. We address two remaining challenges: i) predictor dependence can bias dummy-predictor competition; ii) computation is wasted on recomputing terms shared across experiments. Building on memory-efficient virtual dummies, which sequentially sample projections from their exact conditional law, we estimate the conditional Gaussian law of inactive predictors and map draws onto the remaining sphere radius. A shared lazy Gram cache computes the response product and each requested Gram column once across experiments. Simulations across three covariance structures show that uniform spherical dummies can exceed the target FDR, whereas the proposed method empirically controls FDR. Caching yields more than a sixfold speedup, and the method remains feasible with 100000 predictors, where competing FDR-controlling methods become computationally impractical.

Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During listening, the model can finish processing an audio unit before the next arrives; we term the remaining interval listening-time slack. We propose acoustic-to-text KV compression, which introduces a transcription side channel to convert incoming speech into compact textual memory within this interval. When the cache exceeds a target budget during inference, older acoustic states are evicted while transcripts and recent acoustic context remain. We train the side channel with LoRA using cross-entropy on transcription segments. To preserve listening and speaking behavior, we apply knowledge distillation to the original model's token-level output distributions at native prediction positions. On ten-minute LongSpeech sessions, our MiniCPM-o 4.5 implementation reduces peak streaming KV-cache size by 64.6% compared with the same model without eviction. The proposed method also improves transcription, temporal question answering, and summarization over the baseline. Full-Duplex-Bench evaluations further show comparable pause-handling, turn-taking, and interruption performance.

This article develops a physics-informed deep learning framework for single-snapshot 3D phase-only positioning in narrowband phase-coherent distributed multiple-input multiple-output (D-MIMO) networks under two-ray propagation. Unlike prior phase-coherent D-MIMO localization methods that largely assume line-of-sight (LoS)-only channels, we consider a LoS path and a specular ground reflection. Since the narrowband single-snapshot observation cannot resolve these components in delay, their coherent superposition induces structured perturbations in the carrier phase measurements. To address this, we develop a Gaussian process (GP)-based model that learns the quasi-periodic phase distortion caused by ground reflection from carrier phase measurements collected at a small number of training locations and uses it to generate high-quality synthetic samples. Building on these samples, we propose the Phase-Only Positioning Transformer (POPT), an encoder-only transformer that captures inter-antenna point (AP) phase relationships without solving the highly non-convex maximum-likelihood (ML) problem underlying model-based estimators. We also derive the fundamental position error bound (PEB) and develop maximum-likelihood estimation (MLE) for the considered multipath phase-only D-MIMO positioning problem. Numerical results show that, with only 50 GP training locations, the proposed method achieves near-PEB accuracy. The analytical MLE is highly sensitive to relative permittivity mismatch, whereas the proposed method mitigates this dependence by learning the phase perturbation directly from measurement data. Compared with MLE, the proposed approach also reduces floating-point operation (FLOP) complexity and inference time by about 1.7 and 3.6 orders of magnitude, respectively.

Large-scale Digital Surface Models (DSMs) can be produced cost-effectively from satellite images via stereo-photogrammetry. However, the resulting 3D maps are often contaminated by noise, outliers, and voids. On the other hand, aerial LiDAR provides high-accuracy elevation measurements at a substantially higher cost. In this work, we study diffusion models conditioned both on photogrammetric DSMs and Pléiades imagery to refine vertically co-registered DSMs. We introduce a modified Stable Diffusion 3 architecture with a pruned text stream and a patch-wise normalization strategy, enabling stable training on LiDAR data and transfer from natural images to elevation maps. Experiments in French cities demonstrate that multimodal conditioning improves elevation accuracy, reducing Dense Urban RMSE from 6.00 to 3.45 m in the in-context cities and from 4.16 to 2.77 m in the held-out city of Bordeaux.

Automated EEG artifact removal may improve downstream analysis but can also alter predictive information. We benchmarked nine automated artifact removal methods against a common no-artifact-removal baseline for cross-dataset EEG age prediction. We introduce Signal Quality Index (SQI)-guided GEDAI, which leverages local signal-quality assessment to restrict correction to the channel–epoch pairs requiring intervention. Three deep neural architectures were trained on TUEG and evaluated without target-domain fitting on ds005385, LEMON, and TDBRAIN. Across this setting, GEDAI and SQI-guided GEDAI were the only methods with consistent gains over baseline in age prediction performance across all datasets and architectures ($\Delta$MAE $=-0.77/-0.64$ years, $\Delta R^2=+0.083/+0.072$, respectively). The remaining methods were neutral or detrimental on average ($\Delta$MAE $=+0.22\pm0.16$ years, $\Delta R^2=-0.023\pm0.016$ across methods). The two GEDAI-based methods achieved closely matched performance, while SQI guidance reduced the median modification ratio from $74.78\%$ to $42.80\%$. These findings show that curation benefits are method-dependent and establish SQI guidance as a more selective operating point, leaving more of the original EEG unchanged and limiting the potential loss of neural activity while retaining most of GEDAI's predictive benefit.

Current Audio-Visual LLMs (AVLLMs) struggle with reasoning over videos featuring multi-speaker dialogues. In such videos, resolving "who says what" is crucial, which necessitates trimodal (text-audio-visual) binding. Motivated by these challenges, we systematically investigate how this trimodal binding is achieved in AVLLMs. Specifically, we identify emergent symbolic trimodal binding mechanisms in AVLLMs that utilize modality-specific symbolic variables. By encoding auditory and visual components into symbolic variables-capturing temporal utterance sequences and spatial entity coordinates, respectively-the model establishes cross-modal linking within this abstract space. Crucially, we reveal that when trimodal binding fails, the breakdown predominantly stems from misaligned audio-visual connections. To overcome this bottleneck, we introduce an audio-visual prompting method utilizing an off-the-shelf Active Speaker Detection (ASD) model. By simply overlaying visual bounding boxes on active speakers, this training-free approach yields immediate performance gains across four conversation-centric benchmarks. Moreover, lightweight fine-tuning of fewer than 300 steps on these ASD-prompted-videos extends these gains to three general AV benchmarks, suggesting the generalizability of our method.