Papers

Diabetic retinopathy is a leading cause of preventable blindness; its early lesions are small, low contrast, and easily missed in manual screening. Most automated detectors handle the four non-proliferative DR lesions: microaneurysms, hemorrhages, hard exudates, and soft exudates, with a single multi-class model, even though these lesions differ sharply in size, color, morphology, and prevalence, so a shared model favors common, easy classes over rare, difficult ones. We present PRISM-DR, a lesion-specific pipeline that trains one single-class detector per lesion, each with its own configuration. From a raw fundus image, the pipeline applies region of interest cropping, fundus-specific preprocessing, four parallel YOLO detectors, tiling, per-lesion ensembling of five cross-validation folds, and an inter-lesion suppression step that resolves overlaps by physical lesion size and clinical priority rather than confidence. Per lesion, the best of five YOLO generations is selected, and augmentation is tuned by Bayesian optimization. Trained on IDRiD with stratified five-fold cross-validation, the system reaches a test mAP50 of 0.527 and F1 of 0.529, highest AP50 on hard exudates with 0.561. Without fine-tuning, the models transfer well where the imaging scale is close to IDRiD and degrade as field of view and resolution depart. These modest absolute results reflect a small single-source training set and a difficult task; however, treating each lesion as a separate detection problem is a practical alternative to a single multi-class model.

The transition toward sustainable urban infrastructure is driving a rapid increase in distributed energy resources, particularly among prosumers equipped with solar photovoltaic (PV) installations. Such PV systems are highly dependent on solar radiation, which fluctuates over time and across different city districts, leading to substantial spatial and temporal variations in local carbon intensity factors. The carbon intensity factor is a critical parameter for optimizing energy consumption and reducing emissions in context of smart building control, electric vehicle charging or district heating systems. Hence, high-resolution information on local carbon intensity factors is crucial for intelligent control strategies that minimize environmental impact across urban infrastructures. In this paper, we introduce a novel approach for calculating spatially and temporally resolved carbon intensity factors on an urban district level. The results show that, even within a single city, these factors range from 0 up to 316 g/kWh for an exemplary summer noon. We further develop a transferable surrogate model that estimates these factors using only a limited set of input parameters, typically available to municipalities or grid operators, aligned with census grid cell information. For the surrogate model we compare both traditional decision tree based methods and a state of the art neural network architecture. Our findings demonstrate that the surrogate model can provide highly accurate estimations, particularly in densely built urban areas, supporting data-driven sustainable city operations and indicating transferability even to regions with limited available information. A case study further reveals that neglecting local variations can lead to emission estimation errors of up to 9% for an exemplary building.

Optimal sensor placement is a fundamental problem in graph signal processing (GSP), where a limited number of sensors are deployed to reconstruct a continuous signal field. Existing GSP methods rely on combinatorial optimization over discretized graphs, resulting in high computational cost and sensor locations restricted to graph vertices. We reformulate sensor placement as a continuous-space generative modeling problem. An offline GSP optimization routine first generates training samples of optimal sensor configurations, from which a flow matching (FM) model learns their distribution. At inference, the learned velocity field directly generates continuous sensor coordinates, eliminating online combinatorial optimization. We further develop a permutation-invariant conditional generation framework for deployment with fixed anchor sensors. Experiments on a realistic radio propagation simulator demonstrate the effectiveness of the proposed framework for sensor placement.

Large language model (LLM) agents have shown growing capabilities in tool use, code execution, artifact inspection, and iterative revision, creating new opportunities for automating scientific research. To the best of our knowledge, this paper presents the first end-to-end autoresearch framework for the wireless domain, with a particular focus on wireless resource allocation optimization, an essential area for characterizing the fundamental performance limits of wireless systems and enhancing their practical performance under dynamic channel and network conditions. Specifically, we propose the Wireless AutoResearch Agent (WARA), a closed-loop multi-agent system that transforms an initial research topic into a complete research package. WARA organizes the research workflow into three phases: 1) research gap identification and problem proposal, 2) optimization modeling, algorithm design, and experimentation, and 3) research deliverable construction. Each phase follows an artifact-mediated process, in which structured upstream artifacts are consumed to generate downstream outputs. Controller-managed gates validate these artifacts and maintain consistency among problem formulations, algorithms, experiments, and research claims. When validation fails, WARA repairs only the affected artifact instead of restarting the entire workflow. We further design an LLM-based ScoringAgent to evaluate manuscript-level research validity. Comparative results show that WARA substantially outperforms one-shot LLM generation and approaches the quality profile of recently accepted peer-reviewed papers. These results demonstrate the potential of closed-loop artifact control for end-to-end LLM-assisted wireless optimization research. The source code is available at https://github.com/guoyuan-dotcom/WARA_CUHKSZ

Chronic respiratory diseases affect hundreds of millions of people worldwide, with spirometry serving as the gold standard for pulmonary function assessment. However, conventional spirometry's reliance on mouthpieces and nose clips creates discomfort and technical challenges that can compromise test quality, particularly in pediatric populations where cooperation difficulties are amplified. This study presents SpiRadar, a comprehensive radar-based framework for non-contact spirometry that eliminates physical contact requirements while enabling accurate curve reconstruction, clinical parameter estimation, and bronchodilator response (BDR) assessment. Our key contributions include: (1) a methodological framework integrating physiologically-motivated preprocessing for robust signal extraction, a feature-dependent polynomial transformation linking radar-measured thoracic displacement to spirometric volume curves, and sparse optimization enabling generalization to unseen subjects without individual calibration; (2) a rigorous validation on a pediatric cohort of 39 subjects (ages 6-18 years) undergoing 58 spirometry trials, including healthy children and asthma patients tested pre- and post-bronchodilator, using subject-level leave-one-out cross-validation; (3) clinical-grade performance outperforming alternative non-contact methods, achieving accurate curve reconstruction (mean RMSE: 0.23 +- 0.12 L) and strong correlations (above 0.84) for key spirometric parameters. BDR classification achieved 89.5% accuracy with balanced sensitivity (90.9%) and specificity (87.5%). These results establish clinical feasibility for our non-contact spirometry approach, with particular promise for pediatric asthma monitoring and remote patient care.

A transmission scheme for secure communications, called secret-message transmission by echoing encrypted probes (STEEP), can deliver a positive secrecy rate, even when eavesdropping channels are much stronger than that between users, subject to sufficient asymmetric power allocations from users. This paper considers an optimization problem for STEEP applied to MIMO (or MIMOME) channels between multi-antenna users and multi-antenna Eve. We reveal for the first time a significant increase of the achievable secrecy rate of MIMO STEEP using optimized precoders.

Recent advances in neural rendering have unlocked unprecedented capabilities in 3D reconstruction and novel view synthesis, giving rise to applications such as virtual fly-throughs of a 3D scene reconstructed from a set of sparse, casually captured images. However, these renderings are viewed on a computer screen or conventional VR headsets as 2D images, greatly limiting the perceptual realism and immersiveness of such experiences. The rapid development in novel 3D scene representations calls for dedicated rendering algorithms that convert these readily-available 3D contents into formats that are compatible with emerging 3D display technologies, such as holographic displays. In this paper, we propose a wave-optics rendering pipeline that works with multiplane images (MPIs) for efficient and high-quality hologram synthesis. Our MPI-based computer-generated holography algorithm greatly outperforms state-of-the-art primitive-based CGH algorithms in terms of runtime, achieving speedups up to 250,000x while achieving comparable image quality, and significantly outperforms conventional layer-based CGH algorithms in terms of image quality. We validate our method extensively on a wide variety of 3D scene datasets both in simulation and through experimentally captured results, showing exceptional 3D focal stack and 4D light field reconstruction performance without sacrificing efficiency.

Secure communications over SIMO and MISO channels between wireless nodes are commonly encountered in applications and widely considered in the literature. This paper investigates such a problem by considering a newly proposed scheme called secrecy-message transmission by echoing encrypted probes (STEEP). We focus on a type of heavily eavesdropped channels where an aggressive eavesdropper (Eve) has multiple antennas and is capable of jamming and eavesdropping in the full-duplex mode. Assuming optimal jamming and eavesdropping by Eve, we present the secrecy rates achievable by STEEP in two forms: using SIMO channel in phase 1 and using MISO channel in phase 2, and vice versa. With any given channel condition between every pair of nodes and any finite jamming power by Eve, the secrecy rate of STEEP in either form is shown to be positive subject to a positive power in phase 1 and a sufficiently large power in phase 2. The connections between STEEP in either form and the conventional SIMO and MISO schemes are provided, and their secrecy rates are compared comprehensively subject to random fading channels and finite transmit powers.

We introduce an open-source library for machine-checked control theory in the interactive proof assistant Lean to lay foundations for the verification of cyber-physical systems. To this end, as representative theorems, we present formalizations of Lyapunov stability theory and the small-gain theorem. First, the machinery employed for formalizing Lyapunov stability, i.e., neighborhood filters, allows stating a Lyapunov theorem that covers both points and sets and applies to continuous, discrete, and hybrid systems. Second, the small-gain theorem is proved via stating input-output systems as relations without the usual well-posedness assumption. The Lean formalization of each of these theorems is then presented. We conclude by discussing the library architecture and mentioning some of the other system theoretic results that are formalized in the library along with future plans.

Fully actuated unmanned aerial vehicles (UAVs) are usually certified through rank conditions on a control-allocation matrix or through free-flight tracking performance. For aerial physical interaction, this certification may be incomplete. During sustained contact, part of the available wrench is consumed by the interaction task, and only the residual wrench remains available for stabilization, disturbance rejection, and maneuvering. This paper introduces a control-theoretic framework for contact-persistent full actuation. A rigid-body model on $\R^{3}\times\SO\left(3\right)$ is combined with a morphology-dependent wrench map that captures fixed-tilt, variable-tilt, coaxial, and overactuated multirotor architectures. We define feasible wrench sets under actuator limits, residual wrench sets under task loading, and residual authority margins that strengthen the usual rank-based notion of full actuation. The main result shows that contact-persistent full actuation is equivalent to interiority of the task wrench in the constrained feasible wrench polytope, and that the residual authority radius is exactly the distance to the polytope boundary. We further introduce a signed residual-margin certificate for infeasible and boundary cases, a slack-maximizing allocation certificate, and a robust implementability condition that can be used as a margin-aware safety filter. Numerical evaluation on an abstract tilted hexarotor shows that full row rank alone does not imply feasible contact operation. Intermediate tilt angles preserve residual authority during pushing, whereas small or excessive tilts fail because of lateral-force deficiency or hover-margin loss.

Accurate full-band channel acquisition in frequency-hopping time-division duplex (TDD) systems is challenging because each sounding slot observes only a limited frequency subband, while conventional single-slot recovery cannot fully exploit historical observations. We propose ST-DDA, an online sparse-subspace tracking framework for latest-slot reconstruction in the Doppler–delay–angle (DDA) domain. We first show that the Doppler-domain representation remains energy-concentrated under moderate channel variation, thereby supporting windowed DDA-domain sparse recovery. A local stability analysis further shows that the substantial overlap between adjacent windows enables the preceding-window estimate to warm-start each window-specific recovery problem, allowing the optimization progress to be carried across slots under a fixed per-slot iteration budget. For computational tractability, ST-DDA employs the alternating subspace method, which restricts the regularized least-squares fidelity updates to support-induced subspaces, together with position-encoded convolutional reweighting that exploits local angular and Doppler structures. Experiments show that reweighted ST-DDA achieves more accurate and reliable reconstruction than dynamic compressed-sensing baselines, particularly for longer sounding intervals and larger frequency-hopping periods, while maintaining comparable per-slot runtime.

The growing penetration of distributed energy resources (DERs) has increased the operational variability of distribution networks, making voltage regulation increasingly challenging. Conventional deep reinforcement learning (DRL) methods exhibit unsafe exploration behavior, slow convergence, and limited reliability, which restrict their applicability in safety-critical power system settings. This paper presents a human-interactive reinforcement learning framework that enhances the safety and robustness of autonomous voltage regulation. The proposed approach integrates a Soft Actor-Critic (SAC) agent with an adaptive Lagrange constraint mechanism to enforce voltage limits, while a human-guidance module provides sensitivity-based corrections through coordinated capacitor-bank and Battery Energy Storage System (BESS) dispatch. These corrections are incorporated into a human-regularized actor loss, enabling the policy to internalize safe and interpretable control behavior. The framework is implemented in the PowerGym-OpenDSS environment using the IEEE 13-node feeder. Simulation results show that the proposed Human-Interactive Lagrangian SAC (HI-LSAC) achieves significantly lower voltage-violation severity and reduced power losses compared with the other baseline methods.

This paper proposes a human-on-the-loop resilient control architecture for grid-supporting inverter-based resources (IBRs) operating under actuator degradation. Conventional fault-tolerant control and adaptive control strategies each face notable limitations in this setting: active FTC depends on fast, accurate fault detection and isolation, leaving it vulnerable to misdiagnosis of incipient or ambiguous degradation, while passive FTC tends toward overly conservative operation. Adaptive controllers face a related problem as they typically assume sufficient control authority, but when actuator degradation erodes that authority, the adaptive law may misinterpret tracking errors, leading to parameter drift, performance loss, or instability. To overcome these limitations, the proposed framework embeds human supervisory judgment directly into the control loop, detecting subtle off-nominal behavior, validating or overriding controller parameters, and adjusting operational objectives when conditions exceed the modeled fault space. Two new metrics underpin this resilient decision-making: Generation Reserve Capacity (GRC), which quantifies remaining inverter capacity available for future contingencies, and Controlled Performance Degradation (CPD), which allows temporary, deliberate performance relaxation to preserve overall system operability. A human-selected resilience tuning parameter, {\mu}, governs the trade-off between immediate tracking accuracy and long-term operational readiness. A stability analysis of the proposed HOTL {\mu}-mod adaptive controller is presented. The approach is also validated through simulations of a grid-connected inverter under sequential actuator degradation. Results show that the proposed architecture preserves control reserves, prevents actuator saturation, and achieves superior voltage regulation compared with conventional adaptive control and FTC methods.

Motivated by stochastic model predictive control applications, we develop a semialgebraic framework for constraint tightening in chance-constrained systems with unbounded additive disturbances and saturated inputs. The saturation nonlinearity is represented exactly through its semialgebraic description. A polynomial Lyapunov function satisfying a stochastic drift condition is then synthesized using sum-of-squares optimization, yielding finite-time probabilistic reachable sets and probabilistic ultimate bound. The resulting sets are further optimized through an objective tailored to constraint tightening. Numerical examples illustrate the effectiveness and scalability properties of the approach.

Large-scale AI training turns computing facilities into multi-megawatt loads whose power draw is periodic: tens of thousands of accelerators step in lockstep between compute-bound phases near peak power and communication-bound phases where they idle. Prior work treats each facility as an exogenous periodic forcing on the grid. We pose the operator's question instead: when many independent training jobs share one oversubscribed power envelope, do their cycles stay independent, so aggregate fluctuation grows as the square root of the number of jobs, or can the power-management stack phase-lock them into linear growth? This is emergent synchronization in a population of nonlinear oscillators - the Kuramoto setting - but classical coupling is absent, since accelerator clocks are decoupled from line frequency. We identify the coupling channel in load-dependent throttling: caps, voltage droop, and shared cooling slow computation exactly when aggregate demand is high. Formalizing the fleet as a generalized Kuramoto system, we obtain three operator-facing statements. Dimension: the coupling is repulsive to leading order and turns attractive only when the control loop's phase lag exceeds half a cycle; protection is mode-selective, so rate diversity is required. Detect: frequency-correlated frustration makes the onset first-order and hysteretic. Mitigate: phase-scattering scheduling raises every mode threshold at once. The prediction is falsifiable by a two-job co-capped measurement, which we specify.

Most automatic speaker verification (ASV) systems operate on individual utterances, despite real-world interactions typically consisting of multiple utterances. As speech accumulates, increasingly rich speaker information becomes available through acoustic, prosodic, and linguistic cues, potentially challenging speaker anonymization methods that primarily target vocal characteristics. We investigate ASV in a multi-utterance, multimodal setting and examine whether aggregating information across anonymized speech impacts privacy. We first study audio-only aggregation across multiple anonymized utterances and observe consistent performance improvements as more speech becomes available. We then incorporate prosodic and linguistic information, showing that multimodal systems outperform unimodal approaches. Finally, we compare aggregation strategies and find that frame-level aggregation yields the lowest EERs. Even with only five anonymized utterances, combining audio and text reduces EER by over 15% relative to audio-only aggregation, demonstrating that substantial speaker-discriminative information remains accessible despite anonymization.

Pinching-antenna systems (PASS) reconfigure large-scale propagation environments by activating low-cost radiation points along a dielectric waveguide. However, existing multi-user PASS designs rely on pinching-antenna (PA) position optimization or activation, whose mechanical reconfiguration is much slower than user switching in time-division multiple access systems. To overcome this mismatch, a frequency-tuned beamforming framework (FreqTune-PASS) is proposed, where all PA positions remain fixed and only a user-specific carrier-frequency offset is adjusted in each slot. Since each PA-radiated signal experiences a distinct combination of in-waveguide and free-space delays, a common frequency offset induces path-specific phase rotations and thus reshapes the coherent superposition at the scheduled user. Based on a delay-domain system model, the fast-timescale frequency-offset optimization and the slow-timescale PA-deployment problems are formulated. A closed-form globally optimal offset is derived for the two-PA case, while a low-complexity one-dimensional search and an approximate closed-form offset under equal-spacing deployment are developed for the multi-PA case. The deployment analysis further shows that full-aperture equal spacing maximizes the guaranteed phase-tuning span and yields near-arithmetic relative delays for uniformly distributed users. Numerical results demonstrate that the proposed framework substantially improves both the sum rate and the worst-user rate over fixed-frequency transmission and approaches the coherent-combining upper bound.

In this paper, we propose a hybrid-field channel tracking algorithm for extremely large-scale multiple-input multiple-output (XL-MIMO) systems with mobility, leveraging the historical channel state information. The multiple path scenario is considered, and the line-of-sight path is coarsely estimated through the sparse peak search within a narrow window in the fractional Fourier domain based on the temporal continuity of the user's motion. Moreover, coarse estimations of the non-line-of-sight (NLoS) paths are determined by ensuring the scatterer survival probability, eliminating the necessity of detecting existing NLoS paths. Then, a Newton-based refinement is applied to the estimated paths before seeking possible new paths. Numerical results validate that the proposed algorithm achieves superior performance with low computational complexity.

This paper offers a unique perspective on how speech technologies are being adopted by community-led heritage language preservation and revitalisation initiatives. As a community-led language maintenance strategy, oral histories play a crucial role in Cantonese language revitalisation in New Zealand. The development of Automatic Speech Recognition (ASR) toolkits, such as Whisper, have expedited what has often been a resource and time-intensive process of transcribing oral history collections. However, there is limited research into the effectiveness of ASR toolkits when applied to code-switched language contexts. Based on Word Error Rate (WER), the best performing Whisper model configuration achieved a WER of 12.10 at the expense of accurately transcribing unsupported non-English segments. However, Whisper remains a useful tool by providing a first-pass transcription using only 1% of the estimated time otherwise needed for manual transcription.

This work focuses on training Automatic Target Recognition (ATR) models using simulated Synthetic Aperture Radar (SAR) images to circumvent the lack of real measurements. To obtain robust and versatile ATR models, simulation needs to generate massive datasets that encompass all the variability found in real measurements. Thus, we need a simulator that finds a good tradeoff between execution speed, computational resource consumption, and physical representativeness. In this work, we demonstrate that the Salsa simulator addresses this issue. We ran computing performance tests to show that Salsa can generate 21,600 synthetic images in less than 10 minutes using a single Nvidia GeForce RTX 4090 GPU. Using our ADASCA Deep Learning approach, we demonstrate that these data are sufficiently representative to train ATR models and reach state-of-the-art results on the MSTAR public dataset with an accuracy of 86 %. To illustrate how Salsa unlocks new possibilities to train ATR models, we use the simulator to conduct a study on Electromagnetic (EM) couplings between the targets and their immediate environment. We demonstrate that, if not accounted for in the training dataset, the variability of the EM couplings induced by the variability of the ground surfaces can significantly degrade the performance of ATR models, with an accuracy decrease of more than 4 %. We also show that Salsa can generate in a timely manner (i.e., in less than 4 hours using the same GPU as previously) a massive dataset of 648,000 images with a large variety of couplings to make the ATR models robust to EM coupling variations. Our ATR models can then achieve an accuracy of 87 % on the MSTAR dataset.

This paper investigates how to deploy sensors offline to provide robust passive TDOA localization accuracy across the entire region of interest (ROI) when their positions are subject to drift errors caused by factors such as wind. Since in practice only the 1st and 2nd order statistics of sensor drift errors can be estimated from historical sensor telemetry data or wind field statistics, by using them we first derive a generalized geometric dilution of precision under drift errors ($\mathrm{GDOP_{D}}$), which extends the traditional GDOP ($\mathrm{GDOP_{T}}$). Furthermore, we derive theoretical results related to $\mathrm{GDOP_{D}}$ and $\mathrm{GDOP_{T}}$, revealing that drift errors not only enlarge the value of GDOP but also reshape its distribution, thereby degrading localization performance. Then, we construct a $\operatorname{{GDOP}_{D}}$-based min-max deployment optimization problem. {Finally, we propose an adaptive unidirectional particle swarm optimizer (AUPSO) to solve this challenging problem. The proposed method alleviates the premature convergence and the oscillatory behavior of the traditional PSO. Extensive simulations demonstrate the effectiveness of the proposed method. This research provides a reliable offline sensor deployment planning framework for practical engineering scenarios, when the accurate drift error probability density function is not available.}

We present a goal-agnostic control framework for partial differential equations (PDEs) built around an end-to-end joint-embedding predictive architecture (JEPA). A lightweight 2D vision-transformer (ViT) and action-conditioned latent dynamics are trained offline without a reward or downstream goal, before being frozen and reused by a model-predictive path integral (MPPI) controller. We minimize a control objective in the latent space, initially expressed via the $L^2$ distance and additionally illustrate the benefit of recasting the control objective in terms of an explicit physical observable when available. By instead minimizing the tracking error for a learned linear kinetic-energy (KE) probe on the frozen latent-state rollouts, we demonstrate the ability to reproduce the control of held-out trajectories with $R^2=0.989$, while requiring no change to the underlying world model. For a controlled 2D Navier–Stokes benchmark, using a KE-probe within MPPI planning improves the mean native reward from $-12.08\pm0.86$ for latent-$L^2$ tracking to $-10.90\pm0.91$ (95% CI), all while lowering last-quarter velocity-field RMSE from $0.0765$ to $0.0692$. Across three intentionally withheld, dissimilar, aperiodic targets, KE planning lowers late field RMSE by $53\%$ relative to latent-$L^2$ planning ($0.0220$ versus $0.0469$), winning across 30 paired comparisons. The same frozen model also supports stabilization around a steady-state configuration via direct regulation of KE, achieving $2.7\%$ mean relative error. While the latent probe proves brittle to measurement noise and missing pixels, our findings support the claim that latent dynamics can remain flexible and goal-agnostic, particularly when calibrated observables (granted they guarantee unique continuation) are a suitable objective for state control.

This work examines the deformation dynamics of dielectric liquid droplet when exposed to a uniform electric field. The experimental investigation involves two-phase configurations, here as silicone oil droplets suspended in castor oil. The droplet dynamics including deformation, elongation, oscillation, and rotation are investigated over a range of electric field strengths using shadow-imaging. In parallel, fluorescent tracer particles were employed to perform threedimensional Lagrangian Particle Tracking (3D-LPT) using the Shake-The-Box (STB) technique, allowing for a detailed characterization of the internal electrohydrodynamic flow within the droplet subjected to a uniform electric field. For silicone oil droplets in castor oil medium (S/R > 1), the droplets initially deform into oblate shape. At sufficiently high electric field strengths, the droplet undergoes an Electrohydrodynamic instability, aligning their axis at an angle to the direction of electric field due to the imbalance in induce electric torque. Upon further increasing the field strength, the droplets display oscillatory deformation before settling into a transiently stable configuration. The current study identifies and characterizes the distinct regimes of droplet behaviour, from initial deformation at low electric fields to oscillatory behaviour at high field strengths depending upon the relative dielectric and fluid properties of the liquid phases.

Multi-region unit commitment with reserve sharing requires coordinated optimization across jurisdictionally distinct system operators, exposing sensitive cost curves, topology, and dispatch decisions to inference attacks. The accelerating progress of quantum computing further compounds this threat. As quantum hardware matures, current classically-encrypted data flow becomes vulnerable to retrospective decryption. To enable post-quantum-secure distributed optimization, we propose a customized Benders decomposition-based approach with the global summation structure to share aggregated cuts and variables. By exploiting this structure, we further develop a multi-layer quantum-resilient secure aggregation protocol comprising additive masking for information-theoretic content privacy, affine variable transformation hiding individual sensitive data flows, and reveal-bound lattice-based zero-knowledge proofs providing resilience against active adversaries. Simulation results show that the proposed approach achieves the mean suboptimality of 0.09%-0.22% with lightweight computational overhead, recovers up to 51% of system cost via inter-regional reserve sharing, and imposes no measurable cost-quality trade-off, whereas the noisy ADMM degrades monotonically under tightening privacy budgets and becomes structurally infeasible on combinatorially dense systems.

Quantum state tomography estimates an unknown density operator from measurement data. Dense reconstruction however, can be impractical for many-qubit systems because the Hilbert-space dimension grows exponentially. This contribution develops a rank-adaptive matrix-free approach to low-rank quantum state tomography based on rank-one atomic coordinates. The density operator is represented as a convex combination of pure-state atoms, which preserves positivity and unit trace while avoiding dense density, measurement, and gradient matrices. The resultant algorithm combines atom updates, simplex-constrained coefficient reweighting, and periodic spectral refactorization using only predicted probability vectors, atom vectors, and descriptor-level measurement actions. These computations decompose across measurement outcomes, and admit a master–worker implementation with the measurements partitioned across workers. A rank penalty adapts the representation size during optimization. Provable feasibility, prediction consistency, monotone descent of the penalized objective, and an exact spectral proximal characterization of the refactorization step are established. Simulations with Pauli measurements show favorable accuracy–runtime–memory tradeoffs relative to competing alternatives.

The unit commitment problem determines the optimal strategy to meet the electricity demand at minimum cost by committing power generation units at each point of time. Solving the unit commitment problem gives rise to a challenging optimization problem due to its combinatorial complexity and potentially long solution time requirements. Our proposed solution approach utilizes a decomposition method in conjunction with alternative models from the EGRET library. Results of this decomposition approach tested against four benchmarking systems show that significant computational speed ups are achieved.

Piping and instrumentation diagrams (P&IDs) encode the functional structure of process plants and are a critical yet underutilised source of engineering knowledge for digital twins and intelli-gent decision support. However, digitising legacy P&IDs remains challenging due to heterogene-ous drawing standards and the reliance of existing methods on brittle symbol recognition and rule-based connectivity reconstruction. This work reframes P&ID digitization as the extraction of equipment tags and inference of process topology, rather than graphical reproduction. We pro-pose a two-stage workflow based on multimodal large language models, in which visual extrac-tion and topology reconstruction are treated as distinct reasoning stages guided by chemical en-gineering process knowledge. The approach is evaluated on two ANSI-standard P&ID case stud-ies of increasing complexity. Results show that decomposing visual extraction and topology rea-soning yields more accurate and structurally consistent process representations than end-to-end digitization, highlighting the potential of language-model-based, knowledge-guided workflows for scalable and semantically reliable P&ID digitization.

This paper proposes a human-in-the-loop distributed consensus control approach for demand-side management across multiple buildings. Specifically, a novel framework is introduced in which a human acts as the non-autonomous leader in consensus control of cooperative buildings participating in demand response programs. In this system, the facility manager in the facility building serves as the leader, determining the participation level of cooperative buildings in demand response events while simultaneously considering occupants' comfort. Cooperative buildings align their responses with the facility management building, despite lacking direct access to the facility manager's decisions, which presents a challenge for observer design. To address this, a nonlinear unknown input sliding-mode observer is proposed, tailored for leader-follower multi-agent systems (MASs). Furthermore, a human-in-the-loop leader-follower consensus protocol is introduced, enabling a framework to flexibly manage energy use and balance thermal comfort during demand response events. Simulation results validate the effectiveness of the proposed approach, demonstrating its ability to achieve consensus, maintain system performance, and enhance the adaptability of power grid operations under various demand response scenarios.

Long-video question answering requires a model to preserve visual evidence over time without repeatedly reprocessing the same video. A practical approach is to store the vision-language model's internal key-value (KV) cache for each video chunk and retrieve that state at query time. However, independently cached video chunks do not compose correctly: every chunk is prefilled from local rotary position zero, so naive concatenation collides temporal phases and removes the global order required for questions about what happened first, how often events occurred, or what changed across the video. This paper presents ChronoStitch, a training-free method for composing independently stored visual KV memories. The method first re-bases stored post-rotary keys onto a global three-axis multimodal RoPE coordinate system that preserves time, height, and width structure. We show why a one-dimensional scalar re-indexing is geometrically inconsistent for visual tokens because it turns spatial order within a frame into false temporal displacement. We then address the residual content gap left by positional repair: later chunks were originally encoded without attending to earlier chunks. ChronoStitch therefore selectively recomputes a small fraction of high-deviation later-chunk visual tokens while allowing them to attend over the composed cache. On Qwen2.5-VL-3B and the temporal split of TempCompass, ChronoStitch outperforms naive composition and position-only variants, improving event-ordering accuracy while running 3.3x faster than full joint re-prefilling.

Power grids, markets, and interacting populations, settle into feedback driven equilibria observed through unknown sensors. Our Equilibrium Causal Game (ECG) joins a game to its cyclic causal model, hidden inputs, sensor map, and rules for interventions and equilibrium selection; interventions edit declared objects and recompute equilibrium. Under stated conditions, ECG-separation is sound but incomplete in our examples. Back-door/half-trek routes identify observed queries. Yet for an untouched rotationally symmetric Gaussian block, second moments determine only a source-frame rotation, across which distinct-variable effects generically change. Unknown sensing creates a separate ambiguity. In passive stable linear models without self-effects, unknown wiring and full-rank unknown sensing leave $B$ completely unidentified for $d\ge2$. Under LiNG, non-Gaussianity removes the source rotation; mechanism interventions separate sensing from interactions. With unknown support, invariant sensing, aligned responses, and well-posed single-target interventions identify $(H,B)$ up to declared equivalence. Of $d$ targets, $d-1$ suffice exactly when the sole untargeted node directly parents all others; otherwise $d$ are needed. Acquisition probes are excluded; known wiring gives no universal count. With nonlinear sensing, isotropic Gaussian source blocks admit hidden twists within and across blocks in labelled environments preserving required radial laws. Conversely, under stated positivity, informative one-block changes, rank, and irreducibility conditions, the finest independent source-block representation is identified within the stated alternative class up to block permutation and blockwise coordinate changes, but not downstream mechanisms or the sensor/interaction split. Together, these results show which causal conclusions equilibrium data support and which require targeted experiments.