Societal systems are inherently vulnerable to abrupt, large-scale transformations, including financial crises and trade disruptions, driven by critical transitions that are difficult to predict. Although model-based early-warning indicators have been developed, their effectiveness is often constrained by system-specific assumptions, limiting their applicability to empirical social systems. This study proposes a model-free, data-driven Network Hierarchical Marker (NHM) framework that extracts early-warning signals (EWS) from temporal network hierarchies without requiring explicit system-specific governing equations, parameter identification, supervised training, or predefined crisis labels for the underlying socio-economic dynamics. Applied to international trade networks, NHM identifies warning signals associated with major disruptions in global trade, while revealing vulnerable commodity categories and geographic regions that contribute to systemic fragility. These results suggest that hierarchical network organization can serve as an interpretable structural marker for identifying increasing systemic fragility before or around critical transitions in complex socio-economic systems.
Deploying Federated Learning in multi-UAV networks faces dual challenges: severe resource heterogeneity and non-IID data distributions. Conventional FL frameworks (synchronous or asynchronous) typically address these issues in isolation, resulting in training inefficiencies and communication bottlenecks. To tackle these coupled challenges, we propose AhaFed, an Accelerated Hierarchical Aggregation framework. Specifically, AhaFed integrates: (1) a resource-aware periodic clustering protocol that dynamically groups UAVs to minimize intra-cluster disparities; (2) a hybrid synchronization scheme combining intra-cluster synchronous updates with inter-cluster asynchronous aggregation to mitigate straggler effects; and (3) a delay-aware attention mechanism that weighs updates based on parameter similarity and timeliness to counteract model staleness. Extensive experiments on benchmark datasets demonstrate that AhaFed outperforms baselines, accelerating convergence by 15
The increasing frequency of global wildfires has led to the destruction of vast forests and wetlands. Non-contact remote sensing technologies provide an effective means for accurate burned area segmentation (BAS). However, existing BAS methods often treat each image independently, focusing primarily on local pixel contexts while neglecting the broader semantic consistency of burned regions across different scenes. The lack of global context modeling limits their robustness, as burned areas typically exhibit distinctive and consistent visual characteristics such as color and texture across diverse environments. To address this limitation, we propose a Self-image and Cross-image Consistency Learning (SCCL) framework, which captures both local pixel-level relationships within a single image and global semantic dependencies across multiple images. By enforcing consistent and compact representations of burned regions within and across images, SCCL enhances segmentation robustness under varying weather and terrain conditions. Additionally, to refine boundary delineation between burned and unburned areas, we introduce a Burned Edge Injector (BEI) and an Edge-Injected Decoder (EID). We further construct two large-scale BAS benchmark datasets, BAS-AUS and BAS-EUR, for comprehensive evaluation. Experiments on these benchmarks demonstrate that our method achieves state-of-the-art performance, significantly outperforming previous approaches, with MAE reduced to 0.017 and 0.016, respectively. The new BAS benchmarks and code are available at https://github.com/VisionVerse/SCCL.
Aircraft cabin communication takes on a key role in communication networks. Due to the increasing demand for high data rates among passengers, millimetre-wave (mmWave) and sub-Terahertz (sub-THz) bands, with abundant spectrum resources, are envisioned as promising options. Therefore, detailed channel measurements are required to understand significant channel characterization in aircraft cabins. This paper presents a comprehensive comparison and analysis of channel characterization at 28 GHz, 38 GHz, and 130 GHz based on extensive measurements conducted in an aircraft cabin. A total of 84 transmitter-receiver (Tx-Rx) positions are measured, covering both line-of-sight (LoS) and non-LoS conditions, with Tx-Rx distance ranging from 1 m to 10 m. Based on the measured data, both large-scale and small-scale channel characterization parameters are extracted and analyzed. Statistical models of path loss, shadow fading, Rician K-factor, root-mean-square (RMS) delay spread, and RMS angular spread are proposed. The characterization is analyzed using the power-delay-angular profile and power-angular spectrum. The multipath components are clustered using the density-based spatial clustering of applications with noise algorithm to analyze their intra-cluster delay spread and intra-cluster angle spread. In addition, this paper analyses the system capacity and outage probability, providing some basis for communication system design and planning. To the best of our knowledge, this paper is the first to have both mmWave and sub-THz measurements and analysis performed on an aircraft cabin.
Recent years have witnessed a rapid development of mobile GUI agents powered by large language models (LLMs), which can autonomously execute diverse device-control tasks based on natural language instructions. The increasing accuracy of these agents on standard benchmarks has raised expectations for large-scale real-world deployment, and there are already several commercial agents released and used by early adopters. However, are we really ready for GUI agents integrated into our daily devices as system building blocks? We argue that an important pre-deployment validation is missing to examine whether the agents can maintain their performance under real-world threats. Specifically, unlike existing common benchmarks that are based on simple static app contents (they have to do so to ensure environment consistency between different tests), real-world apps are filled with contents from untrustworthy third parties, such as advertisement emails, user-generated posts and medias, etc. These contents may inevitably appear in the agents' observation space and influence the task execution process. Systematic investigation of this problem is challenging since the real-world app contents are significantly skewed—testing on normal real-world apps usually cannot uncover any potential risk since most app contents are benign. To this end, we introduce a scalable app content instrumentation framework to enable flexible and targeted content modifications within existing applications. Leveraging this framework, we create a test suite comprising both a dynamic task execution environment and a static dataset of challenging GUI states. The dynamic environment encompasses 122 reproducible tasks, and the static dataset consists of over 3,000 scenarios constructed from commercial apps. We perform experiments on both open-source and commercial GUI agents. Our findings reveal that all examined agents can be significantly degraded due to third-party contents, with an average misleading rate of 42.0% and 36.1% in dynamic and static environments respectively. The framework and benchmark has been released at https://agenthazard.github.io.
