This paper presents a low-complexity, model-free, output-feedback controller for a class of unknown time-varying nonlinear systems with unknown input constraints. The controller achieves the preset control accuracy when the actuator is not saturated and maintains flexible control accuracy after actuator saturation. This result extends existing constraint control methods for linear manifolds to a more general form, including the construction of nonlinear manifolds and various types of constraints, thereby achieving preset control accuracy within finite or fixed time. Additionally, flexible control under unknown saturation is achieved through the construction of an error-driven flexible constraint. Finally, second-order and higher-order control examples and simulations are provided.

This paper introduces a systematic method for designing robust linear controllers using output feedback in the presence of operational constraints. The design uses Nagumo's Theorem and the Comparison Lemma to guarantee constraint satisfaction, while incorporating min-norm optimal control principles inspired by Control Barrier Functions. The resulting controller is a continuous piecewise-linear output feedback policy that preserves the closed-loop system's analyzability using linear systems theory. Due to the linear control design, multi-input multi-output (MIMO) robustness margins can be derived with and without active operational constraints. This paper shows that operational constraints on the system's state can be satisfied using an observer-based output feedback control design. Through flight control trade studies, we demonstrate the practical relevance of the framework in safety-critical aircraft control applications.

Distribution networks are transitioning from passive to active systems due to the growing integration of distributed energy resources (DERs). Peer to Peer (P2P) energy trading has emerged as a viable framework that enables local energy exchange among participants, represented here as aggregated microgrids (MGs). Incorporating network constraints is essential to ensure that P2P transactions remain physically feasible and consistent with grid's operating limits. However, existing P2P frameworks still lack advanced predictive mechanisms that allow prosumers to anticipate network feasibility or the distribution system operator (DSO) response during trade formulation. This paper proposes a learning augmented P2P and DSO interface that predicts the DSOs response to the proposed P2P trades, allowing prosumers to self-assess and refine their trading decisions. A supervised transformer based regression model is trained to enable MGs to locally predict the DSOs response without sharing their proposed trades, thereby reducing transaction overhead, alleviating DSO burden, and preserving information privacy. The proposed framework is validated on the modified IEEE 33 bus distribution power system with interconnected microgrids. Case studies are presented to validate the effectiveness of the proposed model in terms of market efficiency, trade acceptance and computational burden.

This study investigates radar technology for non-invasive brain imaging and tumour detection, offering an alternative to MRI and CT scans. Using Ansys HFSS to simulate electromagnetic interactions in brain tissues, we evaluate the penetration, signal strength, and safety of Patch and Vivaldi antennas. Results show Patch antennas are optimal for tumour localization, while Vivaldi antennas suit broader scanning applications. Although promising for safer, more accessible imaging, especially in resource-limited environments, further research with diverse models and actual patient data is essential to advance this technology in non-invasive medical diagnostics.

Eduardo Sontag and coauthors studied Input-to-Output Stability (IOS) and the output asymptotic gain property. These notions changed control theory and recently had an impact on robust adaptive control through the Deadzone-Adapted Disturbance Suppression (DADS) control scheme. Moreover, recently the notion of IOS was extended to systems described by Partial Differential Equations (PDEs). In this work, we celebrate Eduardo Sontag by combining DADS and IOS for PDEs: we study the partial-state regulation problem for a scalar Ordinary Differential Equation (ODE) which is interconnected with a possibly infinite-dimensional system. In such a case the DADS control scheme can allow an escape from the requirements of the small-gain theorem that is mainly used for partial-state feedback. We show the design procedure of partial-state DADS controllers and we prove robust regulation even in the presence of external inputs (disturbances) without assuming knowledge of any disturbance/parameter bounds. The DADS controller is applied to three different cases of the interconnection of an ODE with an almost completely unknown: (a) heat PDE, (b) transport PDE, and (c) wave PDE with viscous damping. We show that the same DADS controller can achieve robust regulation in all three cases.

Automatic subjective speech quality assessment (SSQA) traditionally estimates speech quality on an utterance or system level. While this resolution was adequate for older transmission or synthesis systems that produced speech signals of mediocre quality, modern systems generate high-quality speech with degradations that may occur only locally. With suitable model architectures and regularization losses, SSQA models trained with utterance-level targets can also yield useful local predictions of speech quality. In this work, we extend such models to produce frame-level embeddings that cluster by degradation type. Specifically, we employ a partial mix-up strategy on a parallel corpus of clean and degraded utterances and apply a contrastive loss to distinguish between degradation types. Through experiments on both in- and out-of-domain data, we demonstrate that our approach improves degradation detection and enables the identification of degradation types by analyzing embedding clusters.

Motor-imagery (MI) EEG can be classified using supervised machine learning techniques such as Linear Discriminant Analysis applied to features extracted by Common Spatial Patterns. Performance of these models varies widely, possibly due to MI studies commonly utilising differing post-cue time windows and frequency bands to one another. This study aims to assess how the simultaneous optimisation of both these parameters impact MI classification performance. This is done by iteratively training and testing a series of subject-specific models on different combinations of frequency bandwidth and time window options across 109 subjects. This is followed by a statistical analysis using repeated measures ANOVA to uncover significant differences between different bandwidths and time windows in terms of accuracy across the patient cohort. The resulting visualisations and statistical tests show that there are, indeed, significant differences between both specific time windows and specific bandwidths in terms of accuracy. While the comparison of classification accuracies across 23 frequency bandwidths during five different time windows demonstrates an optimal temporal and spectral scale combination of (0, 4) s at the range of (4, 12) Hz across all subjects, the subjects demonstrate similar accuracies for other parameter combinations. These findings highlight the efficacy of personalised models to detect optimal temporal and spectral parameter combinations to best classify MI EEG signals that inherently vary across subjects.

This paper presents a method to approximate regions of attraction of unknown nonlinear dynamical systems from data. Assuming point-wise evaluations of the vector field and known Lipschitz bounds, a polyhedral uncertainty set of admissible dynamics is constructed. This uncertainty description enables the synthesis of a continuous piece-wise affine Lyapunov candidate via a linear program, enforcing a robust decrease condition for all admissible vector fields. The approach allows certification of a region of attraction consistent with the available data. Numerical examples illustrate the effectiveness of the proposed method in extracting certified regions of attraction from sparse data.

Target speaker extraction (TSE) aims to recover the speech of a desired speaker from a mixture given a short enrollment utterance, while speech enhancement (SE) focuses on improving speech quality under noisy conditions. Most existing TSE and SE systems are based on discriminative modeling and have shown strong interference suppression ability, but they often remain limited in perceptual quality and naturalness. To address this issue, we first introduce LauraTSE, a generative TSE model built on an autoregressive decoder-only language model. Although generative modeling is promising for quality enhancement, purely generative TSE may suffer from hallucination, content drift, and limited controllability in complex acoustic conditions. We therefore propose a discriminative-generative two-stage framework, where a discriminative front-end first produces target-related representations with strong interference suppression, and a generative back-end then reconstructs high-quality speech in the neural audio codec representation space. This design combines the controllability of discriminative extraction with the reconstruction capability of generative modeling. We further investigate several collaboration strategies for the two-stage framework, including front-end freezing, joint fine-tuning, SI-SDR regularization, and autoregressive/non-autoregressive inference. Experimental results on both TSE and SE benchmarks show that the proposed framework achieves a better balance among perceptual quality, intelligibility, and speaker consistency than purely discriminative or purely generative baselines.

Voice anonymization systems aim to protect speaker privacy by obscuring vocal traits while preserving the linguistic content relevant for downstream applications. However, because these linguistic cues remain intact, they can be exploited to identify semantic speech patterns associated with specific speakers. In this work, we present VoxATtack, a novel multimodal de-anonymization model that incorporates both acoustic and textual information to attack anonymization systems. While previous research has focused on refining speaker representations extracted from speech, we show that incorporating textual information with a standard ECAPA-TDNN improves the attacker's performance. Our proposed VoxATtack model employs a dual-branch architecture, with an ECAPA-TDNN processing anonymized speech and a pretrained BERT encoding the transcriptions. Both outputs are projected into embeddings of equal dimensionality and then fused based on confidence weights computed on a per-utterance basis. When evaluating our approach on the VoicePrivacy Attacker Challenge (VPAC) dataset, it outperforms the top-ranking attackers on five out of seven benchmarks, namely B3, B4, B5, T8-5, and T12-5. To further boost performance, we leverage anonymized speech and SpecAugment as augmentation techniques. This enhancement enables VoxATtack to achieve state-of-the-art on all VPAC benchmarks, after scoring 20.6% and 27.2% average equal error rate on T10-2 and T25-1, respectively. Our results demonstrate that incorporating textual information and selective data augmentation reveals critical vulnerabilities in current voice anonymization methods and exposes potential weaknesses in the datasets used to evaluate them.

Speaker anonymization systems hide the identity of speakers while preserving other information such as linguistic content and emotions. To evaluate their privacy benefits, attacks in the form of automatic speaker verification (ASV) systems are employed. In this study, we assess the impact of intra-speaker linguistic content similarity in the attacker training and evaluation datasets, by adapting BERT, a language model, as an ASV system. On the VoicePrivacy Attacker Challenge datasets, our method achieves a mean equal error rate (EER) of 35%, with certain speakers attaining EERs as low as 2%, based solely on the textual content of their utterances. Our explainability study reveals that the system decisions are linked to semantically similar keywords within utterances, stemming from how LibriSpeech is curated. Our study suggests reworking the VoicePrivacy datasets to ensure a fair and unbiased evaluation and challenge the reliance on global EER for privacy evaluations.

The human brain receives nutrients and oxygen through an intricate network of blood vessels. Pathology affecting small vessels, at the mesoscopic scale, represents a critical vulnerability within the cerebral blood supply and can lead to severe conditions, such as Cerebral Small Vessel Diseases. The advent of 7 Tesla MRI systems has enabled the acquisition of higher spatial resolution images, making it possible to visualise such vessels in the brain. However, the lack of publicly available annotated datasets has impeded the development of robust, machine learning-driven segmentation algorithms. To address this, the SMILE-UHURA challenge was organised. This challenge, held in conjunction with the ISBI 2023, in Cartagena de Indias, Colombia, aimed to provide a platform for researchers working on related topics. The SMILE-UHURA challenge addresses the gap in publicly available annotated datasets by providing an annotated dataset of Time-of-Flight angiography acquired with 7T MRI. This dataset was created through a combination of automated pre-segmentation and extensive manual refinement. In this manuscript, sixteen submitted methods and two baseline methods are compared both quantitatively and qualitatively on two different datasets: held-out test MRAs from the same dataset as the training data (with labels kept secret) and a separate 7T ToF MRA dataset where both input volumes and labels are kept secret. The results demonstrate that most of the submitted deep learning methods, trained on the provided training dataset, achieved reliable segmentation performance. Dice scores reached up to 0.838 $\pm$ 0.066 and 0.716 $\pm$ 0.125 on the respective datasets, with an average performance of up to 0.804 $\pm$ 0.15.

A retinal vessel analysis is a procedure that can be used as an assessment of risks to the eye. This work proposes an unsupervised multimodal approach that improves the response of the Frangi filter, enabling automatic vessel segmentation. We propose a filter that computes pixel-level vessel continuity while introducing a local tolerance heuristic to fill in vessel discontinuities produced by the Frangi response. This proposal, called the local-sensitive connectivity filter (LS-CF), is compared against a naive connectivity filter to the baseline thresholded Frangi filter response and to the naive connectivity filter response in combination with the morphological closing and to the current approaches in the literature. The proposal was able to achieve competitive results in a variety of multimodal datasets. It was robust enough to outperform all the state-of-the-art approaches in the literature for the OSIRIX angiographic dataset in terms of accuracy and 4 out of 5 works in the case of the IOSTAR dataset while also outperforming several works in the case of the DRIVE and STARE datasets and 6 out of 10 in the CHASE-DB dataset. For the CHASE-DB, it also outperformed all the state-of-the-art unsupervised methods.

As a rapidly emerging interdisciplinary field that intrinsically integrates microwave and photonics, microwave photonics (MWP) provides disruptive solutions to overcome the fundamental bandwidth of conventional electronic systems. By exploiting the inherently ultra-wide bandwidth and low-loss characteristics of photonic technologies, MWP enables the generation, transmission, processing, and detection of microwave, millimeter-wave, and terahertz signals. Representative breakthroughs include fully photonic microwave radar systems, photonic analog-to-digital converters with bandwidth up to 320 GHz, and photonic wireless communication systems achieving data rate as high as 616 Gbit/s. Meanwhile, the rapid growth of artificial intelligence (AI) is reshaping scientific research, engineering, and daily life in unprecedented ways, such as AI for science/engineering and AI co-scientist/assistant. Correspondingly, AI is profoundly reshaping MWP in all aspects, ranging from signal generation, transmission to signal processing and detection. AI has revolutionized the design, simulation, fabrication, testing, deployment, and maintenance of MWP systems, delivering autonomous operation and exceptional efficiency beyond traditional systems. Motivated by these developments, this Review Paper provides the first comprehensive overview of AI-enabled MWP, systematically summarizing the state-of-the-art advances and presenting insights for both the academic community and the broader public.

In this work we present an efficient and practically implementable approach for the application of reinforcement learning (RL)-based control in chemical process systems. This is an area that has yet to widely adopt RL-based control largely due to inherent challenges in trusting RL algorithms and the time-consuming process of training reliable agents. To address these challenges, we leverage a class of RL algorithms termed Y-wise Affine Neural Network (YANN)- RL, which we have developed in our prior work (Braniff and Tian, 2025a). By strategically initializing actor and critic networks YANN-RL algorithms provide confident and interpretable starting points within control schemes. We apply this RL-based control approach to three different process engineering case studies publicly available on the PC-Gym library (Bloor et al., 2026): (i) a continuous stirred tank reactor (CSTR), (ii) a four-tank system, and (iii) a multistage extraction column. Our approach is compared to several popular RL algorithms (PPO, SAC, DDPG, and TD3) and is benchmarked against nonlinear model predictive control (NMPC). These case studies demonstrate that YANN-RL can greatly reduce the training time and data needed, can be deployed with confidence for chemical process systems, and can approach the performance of NMPC without the knowledge of a full nonlinear model.

Against the backdrop of the burgeoning global low-altitude economy, countries have successively introduced a series of policies to accelerate the application and commercialization of electric vertical take-off and landing (eVTOL) aircraft. Nevertheless, purely electric eVTOLs confront constraints including limited battery energy density, high operational power requirements, and challenges associated with rapid energy replenishment, which collectively restrict their flight endurance and application scenarios. Furthermore, while eVTOL deployment is scaling up, supporting charging infrastructure and regulations remain underdeveloped. This situation presents emerging power distribution networks with new challenges in maintaining adequate electricity supply and ensuring operational continuity. To tackle these issues, following an investigation into battery energy replenishment strategies, a closed-loop supply chain-based model for eVTOL battery charging and swapping is proposed. Time-space network methods are utilized to characterize the scheduling of batteries and logistics throughout the system. Subsequently, aiming to maximize the operational revenue of the model, optimized management of battery swapping, transportation, and charging processes is implemented, facilitating coordinated operation among eVTOLs, swapping stations, and charging stations. Finally, the model is solved by Gurobi, verifying its feasibility. Simulation results further indicate that the model alleviates range anxiety for eVTOLs, offering strong support for their commercialization. Moreover, it enables coordinated scheduling between eVTOLs and the distribution network, thereby facilitating the network's gradual improvement and upgrading.

Semi-blind joint channel estimation and data detection (JCD) is a promising approach to mitigate pilot contamination in cell-free massive multiple-input multiple-output (CF-MaMIMO) networks. The effectiveness of such methods fundamentally depends on identifiability, i.e., the ability to unambiguously recover the unknown channel coefficients and transmitted data signals from the received uplink observations. In this work, we investigate the identifiability of semi-blind JCD from a large-scale system design perspective. We consider a CF-MaMIMO network in which access points (APs) and user equipments (UEs) are spatially distributed according to Poisson point processes (PPPs). The resulting network topology is modeled as bipartite random geometric graph (BRGG) that captures local connectivity induced by wireless propagation. To enable a tractable analysis, the spatially dependent graph model is approximated by a surrogate independent-edge random graph with matched degree distributions. Building on this model, we develop a recursive probabilistic analysis that characterizes the conditions under which semi-blind recovery succeeds with high probability. The proposed analysis reveals an identifiability region as a function of key system parameters, including AP and UE densities and the connectivity radius beyond which channel coefficients are assumed negligible. Monte Carlo simulations validate the predicted identifiability region and assess the accuracy of the proposed graph approximation. The proposed framework provides system level insights into how network density and connectivity affect identifiability in large-scale CF-MaMIMO systems and offers guidelines for selecting deployment parameters and pilot sequence lengths that enable reliable semi-blind recovery.

This letter proposes a network-wide coordinated optimization model to mitigate voltage unbalance (VU) by unleashing the remaining capacity of community inverter-based resources (IBRs). Existing single-sequence strategies ignore coupled capacity constraints and cause idle headroom. Meanwhile, they fail to harness the collective governance capabilities of community IBRs. To solve this discrepancy and exploit the unused potential, we developed a sequence-domain network model in dual commonly shared synchronous reference frames. Strict phase current and apparent power limits are formulated and convexified via polyhedral approximations. A quadratic objective function flexibly balances sequence capacity allocation. Simulation and experimental results validate the effectiveness of the proposed strategy.

We propose a deep beamforming framework for enhancing target speaker(s) in multi-speaker environments. A deep neural network (DNN) is trained to estimate beamforming weights directly from noisy multichannel inputs while satisfying linear spatial constraints through an adaptive multi-term loss inspired by the augmented Lagrangian framework. The loss combines signal reconstruction with penalties that enforce a distortionless response toward the target and suppress the interference subspace. The model is further guided by the target relative transfer function (RTF) and the estimated interference subspace. The proposed model can direct a beam toward the target speaker while directing nulls toward the interfering sources, achieving superior overall enhancement performance compared with the classical LCMV beamformer constructed by the same estimated spatial signatures. Furthermore, compared with the LCMV beamformer, the proposed model produces more controlled sidelobes and improved background-noise attenuation.

Existing LoRaWAN/LoRa simulators consist of large, complicated C++ codebases and often do not support all device classes. This paper presents the design of a simple to use, Python-based discrete-event simulator that addresses these gaps while also introducing a novel method for evaluating real device firmware in the simulator. The simulator is built on a custom asyncio-based simulation kernel, a three-phase packet delivery model that reproduces the capture effect, a full LoRaWAN 1.0.4 stack, and a containerized firmware system that cross-compiles real STM32 C firmware and redirects HAL calls into the simulator via CFFI. The simulator is distributed as a Python package via Github (https://github.com/MatthijsReyers/lora-simulator) and requires no external simulation framework or dependencies.

Reset systems can overcome fundamental limitations of linear time-invariant control. The recently introduced notion of scaled (relative) graphs provides a promising framework for developing graphical analysis and design tools for reset systems, in line with widely adopted loopshaping methods for linear systems. The aim of this paper is to derive techniques for over-bounding the scaled graph of reset systems, and obtain insights in their accuracy. We exploit connections between quadratic dissipativity and scaled graphs to recast the over-bounding problem as the search for piecewise quadratic storage functions. Using specific sampling techniques, we reveal a fundamental limitation of general scaled graph approximation methods that are based on quadratic dissipativity.

Existing Synthetic Aperture Radar (SAR) image generation methods still lack reliable controllability over key imaging parameters, particularly azimuth angle, depression angle, and polarization mode. Our preliminary GeoDiff-SAR supported limited azimuth completion, but remained ineffective for large missing azimuth sectors and did not provide unified control over multiple imaging conditions. To address this problem, we propose GeoDiff-SAR II, a 3D model-guided decoupled framework for controllable SAR image generation. The proposed framework imposes controllability through physically grounded geometric-electromagnetic cues rather than image intensity alone. We introduce a Geometric-Electromagnetic Conditioning Map (GECM), a structured intermediate representation that encodes the target pose map and dominant scattering centers, thereby decoupling macroscopic geometry from microscopic scattering responses. During training, GECMs are derived from real sparse-azimuth SAR images. During inference, the same representation is rendered directly from a 3D CAD model under specified azimuth, depression angle, and polarization conditions, enabling physically consistent control across large viewpoint gaps. The imaging parameters are further converted into text conditions, while the GECM is injected through ControlNet to provide explicit spatial guidance. Combined with Low-Rank Adaptation (LoRA) on a FLUX backbone, the proposed framework unifies geometric-electromagnetic conditioning and parameter-aware generation within a single process. Experiments on simulated and real datasets demonstrate controllable generation over key SAR imaging parameters, stable generalization across large azimuth gaps, and consistent improvements in image fidelity, physical consistency, and downstream Automatic Target Recognition (ATR) performance.

Feedforward steering control is a key component of hierarchical control architectures for autonomous racing. The goal is to reduce steering corrections from the feedback controllers by predicting the vehicle's inverse lateral dynamics. This paper presents a systematic benchmark of two learning-based and two empirical (analytical) feedforward steering controllers. We introduce a new \acf{ehd} formulation based on a polynomial surface fit that captures velocity-dependent nonlinear steering behavior with minimal parametrization. We test the feedforward controllers in a high-fidelity simulation framework based on the real-world Abu Dhabi Autonomous Racing League competition, using a high-fidelity double-track vehicle dynamics simulator. Open-loop evaluation shows that the learning-based controllers achieve the lowest prediction errors; however, closed-loop testing reveals that this improved accuracy does not translate into superior path tracking performance or lap times, even after iterative fine-tuning. In contrast, the proposed EHD approach achieves the best overall closed-loop robustness and lap time, highlighting the necessity of evaluating feedforward strategies within the complete trajectory planning and control software stack. Our code is available at https://github.com/TUMRT/steering_ff_control.

This paper proposes a joint alignment and denoising method for event-based vision sensors (EVSs). Existing signal processing methods for EVSs typically perform event alignment (EA) and event denoising (ED) as separate modules. However, this separation creates a dilemma: without ED, EA is biased by noise, whereas without EA, ED struggles to distinguish signal events from noise ones. To address this dilemma, we jointly optimize EA and ED by formulating a bi-objective Pareto optimization problem. Our formulation is built upon a contrast map that counts the number of events localized in each pixel. With a contrast map, we can formulate EA as maximizing its variance and ED as minimizing the variance. We cast these two conflicting problems as a Pareto optimization and use a regret strategy to obtain a solution. Experimental results on denoising and motion estimation demonstrate that our method achieves improvements against alternative ones.

In this paper, we propose an end-to-end transcoding pipeline, to create 3D Gaussian splatting (3DGS) models from existing 3D plenoptic point cloud or mesh models, when the original multi-view images of the captured 3D object or scene are not available. We also propose a custom initialisation to guide the 3DGS model learning, with constraints to ensure that the final 3DGS model aligns closely with the input point cloud or mesh surface. Tests on a high-quality, standard plenoptic point cloud dataset show that our pipeline produces 3DGS models of high visual quality, with many fewer splats than points in the original dense point clouds. Additionally, our custom initialisation leads to much faster convergence and cleaner surface representation than when starting from the default SfM-based initialisation that is typically used for 3DGS model learning.

Multiple-input multiple-output (MIMO) radar has waveform diversity and large spatial degrees of freedom (DoFs), making it attractive for high-resolution sensing. Scaling MIMO radar to massive arrays can further improve sensing performance, but it also increases hardware cost, power consumption, and digital processing complexity. The microwave linear analog computer (MiLAC) can tackle these challenges by moving linear operations from the digital domain to the analog domain. MiLAC has shown promising benefits for communications in recent studies and this paper identifies its potential for radar sensing. Specifically, we consider both MiLAC-aided transmit beamforming and receiver-side two-dimensional discrete Fourier transform (2D-DFT)-based direction-of-arrival (DoA) estimation. For transmit beamforming, we formulate a weighted Cramer Rao bound (CRB) minimization problem under lossless and reciprocal MiLAC constraints and propose a penalty dual decomposition (PDD)-based iterative algorithm to address the non-convex problem. We further prove that MiLAC-aided and fully-digital beamforming achieve the same CRB. For receiver processing, we show that the 2D DFT can be implemented by a lossless reciprocal MiLAC, which enables analog-domain DoA estimation without digital optimization. Numerical results confirm the theoretical finding and show that the MiLAC-aided approach achieves the same CRB and DoA estimation performance as the fully-digital benchmark. Meanwhile, hardware cost and power consumption are reduced because only low-resolution DACs are required at the transmitter, while RF chains and ADCs are eliminated at the receiver. Moreover, performing the 2D DFT in the analog domain eliminates all digital DFT operations for DoA estimation.

Reasoning has become a defining capability of modern foundation models, yet its development in the audio modality remains limited. Audio poses challenges that are distinct from those of text and vision. It is continuous, temporally dense, and contains linguistic, paralinguistic, and environmental information at multiple time scales. As a result, audio reasoning models must align acoustic signals with the discrete semantic space of large language models, while still preserving fine-grained information needed for reliable inference. Progress is also limited by three major obstacles: the scarcity of genuinely audio-grounded reasoning data, shortcut learning and modality hallucination, and the tension between reasoning depth and real-time latency in spoken interaction. In this paper, we present the first dedicated survey of audio reasoning. We provide a unified formulation that distinguishes direct predictive modeling from reasoning-augmented generation, review the architectural and training foundations of audio reasoning models, and systematically organize recent advances in Audio-to-Text, Audio-to-Speech, Audio-Visual Reasoning and Agentic Audio Reasoning. We further examine emerging paradigms such as Chain-of-Thought prompting, supervised fine-tuning, reinforcement learning, and latency-aware spoken interaction, and discuss evaluation practices, open challenges, and future directions. Our goal is to offer a coherent roadmap for developing robust, efficient, and natively grounded audio reasoning systems.

Most neural video codecs rely on temporal conditioning, which makes them susceptible to error propagation over long sequences. While Transformer-based architectures like the VCT offer a drift-free alternative, they suffer from high computational complexity and inferior RD performance. The recent SWA addresses these shortcomings by reducing complexity and enhancing RD performance, yet it restricts decoding to a strictly sequential raster-scan order, creating a critical bottleneck in decoding latency. To resolve this, we propose P-SWA, utilizing diagonal wavefronts to enable parallel decoding. By embedding a hyperprior and introducing an accumulator to fuse side information and local spatial context, our method increases decoding speed by 36% over the parallel VCT. Simultaneously, it achieves Bjøntegaard Delta-rate savings of up to 10.0% for I-frames and 7.1% for P-frames over the SWA baseline.

Predicting Room Impulse Responses (RIRs) remains a challenge due to the high dimensionality of audio signals and the need for perceptual accuracy. This paper introduces a neural network framework that predicts multi-band Energy Decay Curves (EDCs) directly from room geometry and material properties. Unlike standard models, our framework employs a custom composite loss function that optimizes for both energy levels and decay slopes in the log-domain. This ensures the predicted curves adhere to physical decay principles while maintaining high sensitivity to reverberation time and early reflections. Results demonstrate that the model successfully approximates ground-truth acoustics with minimal error in T30 and clarity indices. The approach offers a computationally efficient alternative to traditional simulations, facilitating realistic audio rendering for interactive virtual environments.

KV cache quantization reduces the memory cost of long-context LLM inference, but introduces approximation error that is typically validated only empirically. Existing systems rely on average-case robustness, with no mechanism to detect or recover from failures at runtime. We present a tiered KV cache architecture that enables runtime-certified attention: INT8 keys and INT4 values are stored in GPU memory, while FP16 originals are retained in system RAM for deterministic fallback. A two-term error decomposition yields per-head, per-step bounds on (i) attention distribution distortion from key quantization and (ii) value reconstruction error. These bounds are computed online and used to drive adaptive precision selection and a multi-stage fallback ladder, which guarantees recovery to the exact dense attention output when required. Across PG-19, NIAH, and RULER benchmarks on LLaMA~3.1-8B with contexts up to 128K, the system matches dense FP16 KV quality within noise for language modelling and retrieval tasks, while recovering catastrophic failures observed in naive INT8/INT4 baselines. Value-sensitive tasks at short context expose a controlled trade-off between compression and fidelity, which can be eliminated via tighter value tolerances or FP16-value fallback. The certification is local (per-head, per-step) and does not guarantee end-to-end model correctness, but ensures that each attention computation is either bounded relative to an FP16 reference or exactly recovered via fallback. This reframes KV cache quantization as a runtime-verified computation rather than a fixed approximation. The goal is not raw speedups, but enabling safe deployment of aggressive KV compression under strict quality constraints.