We study the problem of $(1+\varepsilon)$-approximating the number of four-cycles in graphs given as arbitrary order edge streams. We propose two new algorithms based on sampling induced subgraphs. Our first contribution is a two-pass algorithm that uses $\widetilde{O}(κm / \sqrt{T})$ space, where $m$ is the number of edges, $T$ is the number of four-cycles, and $κ$ is the graph's degeneracy. This algorithm improves upon existing theoretical bounds and is provably optimal for constant-degeneracy graphs, matching the known $Ω(m/\sqrt{T})$ lower bound up to lower-order factors. Our second contribution is a one-pass algorithm that remains accurate when four-cycles are not highly concentrated around individual nodes, edges, or wedges; this structural property is common in sparse social and collaboration networks. We evaluate both algorithms on a variety of real-world graph streams. The two-pass algorithm consistently outperforms state-of-the-art methods, using substantially less space to achieve a desired accuracy. The one-pass algorithm is competitive when four-cycles are evenly distributed, matching our theoretical analysis. Unlike several recent works, our algorithms perform well even on non-bipartite graphs such as social networks.

Evolutionary algorithms have been used for a wide range of multi-objective combinatorial optimization problems. Despite practical success, theoretical results on the runtime of evolutionary algorithms for multi-objective combinatorial problems are rather limited. One classical problem that has been investigated is the multi-objective minimum spanning tree problem for which runtime bounds have been obtained to compute all extremal corner points of the Pareto front. With this paper, we provide some more detailed insights into the structure of the Pareto front when the edge weights take on a small number of distinct values. Based on these insights, we derive new runtime results for evolutionary multi-objective algorithms and complement our theoretical results with experimental investigations.

Generative Retrieval (GR) is an emerging Information Retrieval (IR) paradigm that is motivated by increasingly capable language models. In GR, a model directly generates identifiers for relevant documents. While these systems offer unique advantages, they also introduce distinct failure mechanisms. We explore these failure modes in three contributions: (1) We present a taxonomy of GR failure modes based on GR literature. (2) We empirically investigate failure in a subset of GR: ngram-based methods, more specifically, SEAL and MINDER. Our analysis reveals common issues, such as ambiguous docids, low identifier diversity, and the disproportionate impact of specific identifiers. (3) We introduce a new web-based tool that helps the IR community analyze generated ngrams and their respective contribution to the final ranking, providing an intuitive interface to identify where such GR methods go wrong.

In this paper we tackle the problem of adding varying delay to packets for link emulation. Naive approaches either add more delay than desired or cause packet reordering, both of which are undesirable. We develop adaptive delay correlation, which adds positively correlated delays to packets efficiently. It takes a mean delay and standard deviation (jitter) as input, as well as a half-life period to control the delay dynamics. We investigate the accuracy and dynamics of the resulting packet delays with and without bandwidth limitation. As a result we give a recommendation for the configuration of the half-life period. We implement adaptive delay correlation in a DPDK-based packet delayer and spacer (DPDS), investigate its performance on hardware, and compare it with the widely used link emulator NetEm and the recently developed DPDK-based emulator MoonEm. DPDS outperforms both of them with a zero-loss throughput of 95 Gbit/s for constant delay and, with spacing enabled, 85 Gbit/s for varying delay with 3 ms jitter. Further, DPDS supports packet reordering with zero-loss throughputs of 73 Gbit/s and 58 Gbit/s for constant and varying delay, respectively, as well as policing and two packet loss models.

Recommender systems alleviate information overload, yet repeated feedback between recommendations and user interactions can reinforce existing preferences and narrow users' exposure, forming information cocoons. While this phenomenon has been widely studied in traditional sequential recommendation, its impact on generative recommendation remains unclear. By replacing atomic item IDs with Semantic ID (SID) sequences, generative recommenders introduce a different recommendation mechanism whose role in information cocoon formation is not yet understood. To investigate whether generative recommenders deepen information cocoons, we propose \textsc{RecLoop}, a closed-loop simulation framework with LLM-driven user agents. We compare two generative recommenders and two traditional sequential baselines on two Amazon datasets across multiple feedback cycles. In addition to standard exposure-level metrics, we introduce \emph{Code-Space Structural Cocoon}, a model-level metric that measures concentration in the generated SID space. Experimental results show that generative recommenders are generally less prone to exposure-level cocoon formation than traditional baselines, preserving broader exposure diversity and slowing cross-user homogenization. However, feedback loops can still induce concentration within the generated SID space. We further find that cocoon severity depends strongly on tokenization strategy and model scale: collaborative-signal tokenization produces stronger cocoon effects than semantic tokenization, whereas larger models maintain greater code-space diversity and better retain access to niche content. These findings suggest that information cocoons in generative recommendation are shaped not only by recommendation behavior, but also by item tokenization and model capacity. Our code is available at https://github.com/Dregen-Yor/RecLoop.

The total biharmonic distance, which is the sum of the biharmonic distance between every pair of nodes in a network, is a key metric for evaluating network connectivity and robustness. In this paper, we study the problem of minimizing the total biharmonic distance by adding $k$ nonexistent edges for a given graph $G$ and budget $k$. The problem is computationally challenging. We show that the objective function of the problem is monotone but not supermodular. To solve this problem, we propose simple greedy algorithms with cubic time complexity. To mitigate the high time complexity of these greedy algorithms, we apply several techniques, including the projection method, the Laplacian solver, and convex hull approximation. These techniques reduce the time complexity of our proposed algorithms from cubic to nearly linear while providing error guarantees. Finally, extensive experiments on real datasets demonstrate both the efficiency and effectiveness of our proposed algorithms.

We show that existing tools for termination analysis are extremely well suited for LTL model checking of infinite state systems. To this end, we present a framework MoAT which uses the well-known automata-based approach and reduces the LTL model checking problem to fair termination. To prove or disprove fair termination, it then calls the termination tools KoAT and LoAT in the backend. Our experiments show that in this way, MoAT is on par with existing state-of-the-art tools for LTL model checking of infinite state systems.

5G is a fundamental technology for the digitalization of smart manufacturing. Smart manufacturing relies on the use of digital models to optimize industrial processes before implementation on the manufacturing plants. These models should account for the impact of 5G communications to adequately dimension and optimize 5G-based industrial processes. This paper presents the first integration of industrial digital models with a 5G digital model, implemented as an Asset Administration Shell (AAS) of a 5G system. The two models are interconnected using an OPC UA-based interface. We evaluate the impact of the integrated model using a use case where Automated Guided Vehicles (AGVs) transport material from a warehouse to production lines. The AGVs periodically exchange their positions over 5G to avoid potential collisions. If the communications fail, the AGVs stop for safety reasons until a reliable 5G connection can be guaranteed. We demonstrate that, by integrating 5G and industrial digital models, it is possible to account for, and quantify, the impact of 5G communications on the operation and productivity of industrial processes. This result highlights the importance and necessity of integrating 5G into industrial digital models for their joint design and optimization.

Teleoperated driving (ToD) enables the remote driving or control of vehicles. For this purpose, vehicles must transmit video feeds to the ToD control center so that the remote operator is fully aware of the driving conditions and can safely control the vehicle. 5G (and beyond) networks are fundamental for the deployment of ToD as they can provide the low latency, reliable and broadband connection necessary to connect the vehicle and ToD control center. However, it is unclear whether common 5G network architectures and configurations are well-suited to support the simultaneous teleoperation of multiple vehicles with demanding uplink bandwidth, as current networks are mainly configured to support mobile broadband services. This paper demonstrates that MEC or edge-based 5G networks are better suited to support and scale the ToD service than centralized networks, and quantifies the bandwidth required to simultaneously teleoperate multiple vehicles under various 5G network architectures and configurations, including different duplexing modes and TDD frame structures. Finally, the study shows that the configuration of the control channels can help mitigate the impact that the processing time of the video feeds has on the capacity to support and scale the ToD service.

Future wireless networks must enhance their capacity to sustain deterministic service levels and support emerging time-sensitive services in key verticals. The ability to guarantee bounded latencies heavily depends on efficient radio resource management. Configured Grant (CG) scheduling can reduce latency by pre-allocating resources, but its effectiveness and efficiency decrease under variable traffic patterns. This study presents a novel predictive CG scheduling scheme that pre-allocates resources based on traffic predictions while accounting for prediction inaccuracies. By considering these inaccuracies, the scheme significantly improves the ability to meet bounded latency requirements, which are essential for supporting deterministic service levels. Our evaluations show that the proposed scheme significantly enhances the capacity to support deterministic service levels while improving resource utilization, even in scenarios with variable and mixed traffic flows with diverse requirements.

Due to educational inequality, high-quality lesson plans often mismatch the needs of disparate educational contexts. Teachers typically modify existing lesson plans to fit new contexts, but current tools instead focus on generating content from scratch, creating additional workload. Moreover, a critical gap remains in supporting teachers to quickly adapt to new learning profiles. To bridge these gaps, we present AdaPT, a system leverages LLMs to support transformation of existing lesson plans for cross-regional and differentiated instruction. AdaPT features an interactive interface that allows teachers to input student profiles, offers structured lesson representation, provides explanations for lesson-plan transformations, automatically adapts lesson content for new contexts, and supports iterative, teacher-in-the-loop refinement. We evaluated AdaPT through a user study with 9 teachers and an expert evaluation with 3 specialists. Results show that AdaPT supports workflows of teachers and offers a promising pathway toward promoting educational equity.

Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease, often affecting speech due to bulbar dysfunction. In this study, we predict speech impairment in people with ALS (pwALS) using two clinical speech-related scores. We evaluate cross-sectional (across speakers) and personalised (within-speaker) modelling paradigms and analyse the utility of common speech tasks to contribute to the standardisation of speech data collection for pwALS. Experiments on a German-speaking cohort of 66 pwALS show that repetition tasks (/da/-/da/, /da/-/ba/) achieved the best cross-sectional performance (Concordance Correlation Coefficient (CCC) = 0.62) for predicting the Quality of Life in the Dysarthric Speaker questionnaire, while the within-speaker setting reached a CCC of 0.86. This study represents an initial step towards speech impairment prediction in German-speaking pwALS and highlights the potential of automated speech analysis as a supportive tool for speech impairment assessment.

State-of-the-art traffic flow forecasting models, including Graph Convolutional Networks and graph-less MLPs, require centralized GPU training across all sensors, making them impractical for resource-constrained intelligent transportation deployments. We show that much of this complexity is unnecessary. A parametric analysis of the recent graph-less model GLMST reveals that reducing its internal embedding dimension from 64 to 4 degrades MAPE by less than one percentage point, suggesting that the model's effective capacity far exceeds what the task requires. Motivated by this finding, we replace the neural architecture entirely with per-sensor Ridge regression using horizon-aligned periodic features, combined with Recursive Least Squares (RLS) for online adaptation. With only 444 parameters per sensor (80x fewer than GLMST) and test-time online adaptation, our method achieves the best MAPE on three of four PEMS benchmarks, and remains within one percentage point on the fourth. Because each sensor's model is self-contained and involves only elementary linear algebra, the entire pipeline (training, inference, and online adaptation) runs on edge hardware without a GPU. An ESP32 microcontroller (160 MHz, 520 KB SRAM) completes cold-start training in 7.4s and each predict-and-update in under 2ms with zero heap allocation; a single Raspberry Pi 5 core completes cold-start training in 0.21s and each predict-and-update in 0.26ms.

As software systems grow in scale and complexity, debugging and repair remain costly and time-consuming. Large language models (LLMs) have advanced automated program repair (APR), but existing LLM-based APR approaches still largely rely on static or retrieved context, error messages, and coarse-grained validation outcomes. As a result, they underutilize dynamic information for failure understanding and repair, including failure-execution dynamics and patch-validation dynamics. Effectively leveraging such information, however, is challenging: failure-execution traces are large and noisy, raw static-dynamic context is not self-explanatory, and patch-validation dynamics are often reduced to coarse feedback. To address these challenges, we propose \textsc{PracRepair}, a fully automated LLM-based APR framework inspired by human-like debugging practices. \textsc{PracRepair} constructs an on-demand static-dynamic context from buggy programs and failure executions, performs question-driven failure diagnosis to formulate explicit repair hypotheses, and iteratively refines candidate patches using validation diagnostics and trace-level behavioral changes. Experimental results on Defects4J V1.2 and V2.0 show that \textsc{PracRepair} consistently outperforms state-of-the-art baselines. Specifically, under GPT-3.5, \textsc{PracRepair} correctly fixes 139/136 bugs on Defects4J V1.2/V2.0, while under GPT-4o it further improves to 162/171. Moreover, \textsc{PracRepair} generalizes effectively to RWB (Real-World Bugs), achieving the best performance across multiple foundation models.

The Financial Times 50 (FT50) journal list shapes hiring, promotion, accreditation, and research evaluation across business schools worldwide. Yet journals on the list are typically treated as if they represent a homogeneous tier of excellence. We test this assumption by comparing 53 FT50 and recently removed journals across three distinct impact channels: scholarly influence (field-weighted citations and visibility), policy uptake, and technological reach through patent citations. Using a panel of more than 60,000 publications from 2005 to 2019, we find striking heterogeneity hidden beneath the binary FT50 label. Elite economics journals dominate policy influence, information systems and marketing journals lead technological impact, while many highly cited management journals exhibit limited reach beyond academia. Citation, policy, and patent indicators behave as largely independent dimensions of impact, with a citation-only ranking correlating only moderately with a multidimensional ranking. Nearly half of all journals change quartile once policy and patent indicators are incorporated, demonstrating that assessments based solely on scholarly citations overlook important dimensions of research influence. While the FT50 remains widely used as a binary classification of journal quality, our results reveal a substantial within-list impact spectrum and show that journal rankings are highly sensitive to how impact is defined and measured.

Credibility of a simulation model is an important topic. Several approaches try to quantify the credibility of simulation. However, models are mostly assembled within a simulation architecture. Can the credibility of a simulation architecture be assessed based on the credibility of the models that comprise it? This paper aims to address this issue by providing an overview of the current state of the art in the field of assembly credibility. It will compare sensitivity analysis techniques, qualitative analysis by experts, explainability in AI, and networks. Finally, an assessment of the proposed approaches, based on criteria such as rigor, generalization, and resource requirements, will reveal the strengths and weaknesses of each approach.

We study the problem of data repair, a key task in data cleaning that corrects erroneous entries in raw datasets to improve overall data quality. Although recent data-driven methods, especially those based on large language models (LLMs), achieve remarkable performance, we observe that: (i) they directly repair data in the raw and low-quality context, which may compromise learning signals, and (ii) they directly use uncertain model outputs as repairs, potentially introducing unreliable corrections and compromising repair quality. Motivated by the efficiency of small language models (SLMs) and the capabilities of LLMs, and aiming to address the above limitations, we propose LasRepair, a framework that collaborates Large and small language models for data repair. LasRepair employs an LLM as an instructor, which selects a global repair context to guide the SLM. The SLM acts as a corrector, using the selected context to repair erroneous data more efficiently. Moreover, to further improve context quality, we extend LasRepair to LasRepair+, which formulates data repair as an Expectation-Maximisation (EM) procedure that alternates between an E-step for updating the corrector parameters and an M-step for refining the repair context. Furthermore, to mitigate model uncertainty, we propose LasRepair++, which uses column-calibrated model confidence to down-weight unreliable repaired rows when updating the corrector, thereby enhancing repair quality. Theoretical analysis and empirical evaluation demonstrate the superiority of our methods. We theoretically prove the effectiveness of the EM-style procedure and the confidence-based weighting. Experiments on real-world datasets show that LasRepair++~ achieves an average F1-score improvement of 18.1% over the strongest baseline.

We study exact exponential-time algorithms for Edge Deletion to Cactus. Given a connected graph $G$, the task is to delete a minimum number of edges so that the remaining spanning graph is a connected cactus. Akhtar and Philip (IWOCA 2026) gave an $O^*(3^n)$-time algorithm for the unweighted problem, where $n$ is the number of vertices in the input graph and the $O^*(\cdot)$ notation hides polynomial factors. We improve this bound to $O^*(2^n)$ time and space. More generally, if the deletion costs take at most $q$ distinct nonnegative real values, then the weighted problem can be solved in $O^*(2^n n^{O(q)})$ time and space. Thus every fixed number of distinct costs, and in particular the unweighted case, admits a faster exact algorithm. For nonnegative integer costs of total weight $W$, we obtain an $O^*(2^n(W+1))$ pseudo-polynomial algorithm, while arbitrary nonnegative real costs admit an $O^*(3^n)$ exact algorithm.

This study investigates the small-signal stability of centralized energy storage systems (CESSs) using grid-following (GFL) and grid-forming (GFM) controls, particularly focusing on bidirectional power flow and multiple energy storage systems (ESSs). To address the issue of complex dynamics in CESSs when comprehensive GFL and GFM control loops are considered, high-order dynamics are simplified using the virtual damping method by focusing on the dominant oscillation mode. Damping analysis verifies that CESSs using a single-type control (either GFL or GFM) have dynamic superimposition characteristics. Specifically, as ESS number increases, the damping of GFM-CESSs improves but that of GFL-CESSs decreases. The damping sensitivity shows that the damping of GFM-CESSs is more sensitive to bidirectional power flow and all control loops, whereas that of GFL-CESSs is more sensitive to d-axis control loop. Consequently, GFM-CESSs are preferred for large-scale integration but are limited in scenarios with significant power reversal. If GFL and GFM controls are hybridized in CESSs, the ratio of GFM-CESSs should be constrained to avoid instability from modal resonance between GFL-CESSs and GFM-CESSs. This highlights that implementing GFM-CESSs necessitates considering scenario limitations rather than pursuing maximal integration under hybrid integration conditions. The conclusions are validated through modal analysis and time-domain simulations.

Tool-using LLM agents are shifting the unit of computation from explicit human-issued commands to model-driven tasks with stateful consequences. Yet today's agent runtimes still expose tools as isolated RPCs. This interface gives runtimes a convenient integration point, but it lacks a task-scoped execution boundary for commit, rollback, recovery, and audit across multi-step agent workflows. We argue that this mismatch calls for a runtime containment boundary rather than another per-call guardrail. This paper introduces Cordon, a transactional runtime system for staging and validating irreversible agent effects before commit. A semantic transaction is a task-level execution boundary that binds tool intents and runtime-tracked result lineage to reversible local state, staged external effects, delegated authority, and audit metadata. Cordon implements this abstraction with a transaction manager that tracks derived result objects, executes reversible mutations in shadow state, stages outward-facing actions in an effect outbox, and records recovery metadata. The runtime then validates the composed execution flow before it commits state or releases external effects. Our evaluation across adversarial and benign workflows shows that Cordon exposes cross-step violations missed by existing defenses. It also reduces irreversible-effect failures while preserving benign task completion with modest approval and latency overhead.

It has been well known that inverter-based resources (IBRs) can lead to converter-driven stability issues under weak grid connections. However, as the number of IBRs increases, instabilities can also occur even under strong grid connections. A practical case is presented to demonstrate this conclusion, using large-scale energy storage systems (ESSs) as an example. In this study, the ESSs induce oscillations with a frequency of 150 Hz in the d-q coordinates while providing both capacitive and inductive reactive power support (achieved by ESS functional control loops) to the connected power system. Theoretical analysis reveals that under strong grid connections, the dynamic interactions among power conversion systems (PCSs) of ESSs can be superimposed and intensified as the ESS scale extends, which reduces oscillation damping and leads to system instability. This indicates that ESS functional control loops also have potential instability risks when providing supports to power systems, which should be carefully examined. Finally, major impact factors are identified to mitigate the oscillations, and the conclusions are validated based on the SIMULINK platform. This paper provides valuable practical insights into system instabilities even under strong grid conditions, emphasizing the importance of functional control design and careful planning of the scale for IBR-dominated systems.

Massive controlled DC resources (CDCRs), such as battery energy storage systems, are connected to AC power systems through bidirectional inverters for power balance requirements. This study investigates converter-driven stability (CDS) issues in the sub-synchronous frequency range caused by large-scale bidirectional inverter-based stations (IBSs). The impacts of the AC and DC connections of IBSs on subsynchronous oscillations (SSOs) are compared by examining three factors: the number of CDCRs, power flow direction, and control parameters of the inverters. For AC connections, IBSs may induce instability as the number of CDCRs increases, regardless of the power flow direction. To maintain stability, the maximum power amplitude of the IBS is calculated. It is found that switching to DC connections can reduce these instability risks if the DC line resistance is much less than the AC line reactance. Moreover, the method of tuning control parameters is demonstrated to be more effective in improving power-related critical stability under DC connections. Therefore, The DC-IBS is preferred for high-voltage transmission. Finally, the conclusions are validated in power systems connected with both AC- and DC-IBSs under various network topologies and system scales.

LiDAR sensors are widely deployed in autonomous systems for 3D perception and safety-critical decision-making. We identify a previously unexplored attack surface in which dormant malware embedded in the LiDAR sensing pipeline remains inactive during normal operation and can be externally triggered after deployment, without requiring access to sensor hardware or networking at attack time. To operationalize this threat, we design malware capable of low-level point-cloud manipulation and embed it into LiDAR firmware. This malware was developed in a closed research test environment with vendor technical support, rather than by exploiting an inherent production supply-chain vulnerability. To selectively trigger attack activation, we design and implement an optical trigger that remotely activates the malware by delivering a modulated signal into the sensing environment. Once triggered, the malware performs real-time point cloud manipulation, and we demonstrate false object injection and real object suppression on static and mobile victim platforms. Our evaluation first establishes attack feasibility, including static operation at 300~ft and recorded drive-by runs reaching 35~mph. We then illustrate quantitatively that injected person-like artifacts can remain semantically detectable by a state-of-the-art 3D object detector. Finally, we demonstrate multiple modes of safety-critical impact on a deployed tactical autonomous vehicle. Together, these results highlight the need for stronger integrity guarantees throughout the LiDAR sensor development and deployment pipeline.

Snowpark enables data engineering and AI/ML workloads in Snowflake by executing user-defined functions in secure sandboxes. Many of these workloads require external connectivity to access cloud APIs, external databases, or feature stores, creating a dependability challenge: how to provide transparent network access while preserving strict multi-tenant isolation and resource fairness. This paper presents Secure Network Access in Snowpark (SNAS), a production architecture for secure external communication from sandboxed workloads. SNAS combines Extended Berkeley Packet Filter (eBPF) packet filtering, Generic Network Virtualization Encapsulation (GENEVE) overlay networks, and distributed egress proxies for policy-driven egress control with low overhead. We describe the design, deployment, and measured production behavior of SNAS, including an eBPF-based bandwidth limiter using the Earliest Departure Time (EDT) algorithm, dual-tier policy enforcement, and safeguards for connection limiting and port exhaustion. SNAS is deployed across all Snowflake regions and supports large-scale production workloads including petabyte-scale data transfer and latency-sensitive external integrations.

Projector photometric compensation corrects color distortions introduced by surface texture, reflection, and ambient lighting. Existing deep learning-based methods usually require professional scene-specific data collection and lack consideration for perceptual quality. To address this limitation, we present a diffusion-based photometric compensation method that reconstructs compensation images under photometric and content-aware guidance. Specifically, we first model the photometric distortions introduced during projection as environment-dependent additive noise, thereby reformulating the photometric compensation problem as a denoising task with physical constraints. Next, we introduce a diffusion model, which generates compensation images by following an additive trajectory to iteratively remove the noise. Finally, to accurately estimate the noise at each timestep, by analyzing the factors that contribute to distortions in the physical process of projection and capturing, we design a noise estimation network that incorporates features of both photometry-aware and content conditions. Experiments show that our method achieves superior visual performance in unknown scenarios, thereby exhibiting significant practical advantages over prior methods.

Agentic kernel optimization automates manual GPU kernel tuning via iterative generation, validation, and profiling with reasoning LLMs, casting the optimization task as feedback-guided search. However, our workload characterization reveals three system-level inefficiencies that limit search efficiency: (1) long generation latency due to LLM reasoning, (2) insufficient profiling feedback, and (3) underutilized validation/profiling resources. Our key insight is that the ongoing reasoning generation exposes a window for producing additional candidate kernels before it completes, allowing the system to terminate reasoning early once a satisfactory kernel appears. We present SpecGen, an agentic kernel optimization system with \emph{speculative generation}. First, SpecGen forks non-reasoning generations at well-chosen trigger points in the reasoning trace to yield kernels, increasing the candidate kernel count per iteration. These kernels are validated and profiled in parallel with the ongoing reasoning, increasing profiling feedback, and keeping resources busy during generation. When a kernel meets the termination criterion, SpecGen terminates the reasoning generation early to reduce the generation latency. Second, SpecGen dynamically reallocates validation and profiling GPU pools based on the arrival rate and prioritizes requests to reduce profiling feedback latency under bursty speculative generation load. Furthermore, SpecGen utilizes spare memory of the validation/profiling GPUs as remote KV cache storage to eliminate prefix recomputation of speculative generations under limited memory budget. Experiments with two reasoning LLMs on H200 show that SpecGen reduces end-to-end time over three baseline systems, while producing more profiling feedback, increasing resource utilization, and improving kernel speedup under a fixed time and token budget.

AI standardized patients are becoming a setting for professional training in clinical consultation. This paper presents MedEasy, a multi-agent system that organizes virtual-patient practice through patient dialogue, clinical actions, decision submission, documentation, and feedback. We first conducted a formative study with 12 clinical-year medical students through interviews and three co-design workshops. The findings informed a staged workflow, structured case records, action-contingent findings, and trajectory-based review. We then conducted an evaluative user study with a separate cohort of 12 clinical-year medical students, with each participant completing two counterbalanced cases. Learners interpreted MedEasy as a connected consultation environment. They used patient responses, examination findings, available actions, and feedback together to judge whether the represented case remained coherent. They valued repeatable practice and recorded review, while questioning missing actions and feedback criteria. The paper contributes design implications for AI-supported professional training systems that use case-specific standards to connect situated practice.

Despite rapid advances in UAV technologies, current deployments remain limited due to several gaps in UAV systems research. To address these challenges, we propose OmniDroneX, a unified Drone-as-a-Service ecosystem, in which drones are transitioned from fixed function platforms into dynamically composable entities that can be integrated with external infrastructures to offer omni-capabilities. OmniDroneX bridges low-level physical primitives with high-level mission intent through a unified vendor-agnostic interface (libUAV) and a formal physical-service abstraction model (PT-SOA). A core innovation is the diverse application of large language models (LLMs) across multiple layers of the OmniDroneX architecture. LLMs are used to assist in identifying and formalizing primitive device functions and abstract service definitions, supporting automated service composition and workflow generation, and enabling interactive, natural-language mission specification and refinement. OmniDroneX also incorporates important categories of composition techniques that are essential in dynamic UAV systems, including physical layer composition for drone capability augmentation, as well as spatiotemporal, functional, collaborative, exception-aware, and QoS-based service compositions. Collectively, these features allow OmniDroneX to serve as a foundation for scalable, resilient, and self-evolving UAV ecosystems operating in complex and dynamic environments.

We propose a direct data-driven method for controller synthesis of infinite networks composed of unknown linear time-invariant subsystems. Using a single set of noise-corrupted input-state trajectories collected from each subsystem, and provided that certain linear matrix inequalities hold, each subsystem is rendered exponentially input-to-state stable (eISS) by locally constructing an eISS control Lyapunov function together with an exponentially input-to-state stabilizing feedback controller. We then compose these local components under a compositional small-gain condition in infinite-dimensional spaces to obtain a global control Lyapunov function and an associated stabilizing controller, ensuring uniform global exponential stability of the infinite network. The approach is validated on a physical case study with unknown dynamics.

Reliable, low-latency multi-hop data delivery in vehicular networks is increasingly demanded, yet remains challenging due to frequent route failures caused by high mobility and intermittent blockage. While redundancy-based routing enhances robustness by forwarding packets over multiple paths, over-replication intensifies contention and introduces additional delay, highlighting the need to carefully managing redundancy--reliability trade-off. However, conventional deterministic multi-path replication typically duplicates packets to an integer number of branches, making the redundancy level hard to tune and adapt to time-varying network dynamics in vehicular networks. To this end, Redundancy-Controlled Stochastic (RATIO) routing is proposed in this paper. For each active flow, RATIO constructs a weighted reduced directed acyclic graph (DAG) as the routing structure, where edge weights specify per-link forwarding probabilities. At fork nodes, the aggregate outgoing forwarding probability is allowed to exceed one and a modulo-based stochastic forwarding rule is employed to guarantee feasible forwarding, thereby enabling continuously controllable redundancy. An idealized RATIO design is formulated as a load-minimizing optimization subject to per-flow timely-reliability and link-capacity constraints, but the problem is generally intractable under time-varying wireless dynamics. Accordingly, a practical heuristic, termed H-RATIO, is developed. H-RATIO constructs a compact reduced DAG by taking the union of candidate paths and optimizes forwarding probabilities via local scoring and replication-adjustment iterations. Extensive trace-driven SUMO/ns-3 co-simulations demonstrate that RATIO/H-RATIO consistently achieves the highest timely PDR compared to baselines, while providing substantially better delivery efficiency, especially under high-load scenarios.