Abstract
Municipal meeting minutes are official documents of local governance that exhibit heterogeneous formats and writing styles. Effective information retrieval (IR) requires identifying metadata such as meeting number, date, location, participants, and start/end times, elements that are rarely standardized or easily extracted automatically. Existing named entity recognition (NER) models are ill-suited to this task, as they are not adapted to such domain-specific categories. In this paper, we propose a two-stage pipeline for metadata extraction from municipal minutes. First, a question-answering (QA) model identifies the opening and closing text segments containing metadata. Transformer-based models (BERTimbau and XLM-RoBERTa with and without a CRF layer) are then applied for fine-grained entity extraction, with deslexicalization explored as an additional modeling strategy. We benchmark the pipeline against open and closed-weight LLMs (Phi and Gemini), considering performance, inference cost, and carbon footprint. Our results demonstrate strong in-domain performance, outperforming the evaluated LLMs. Differences observed in cross-municipality evaluation highlight the linguistic diversity and structural variation across municipal records, underscoring the challenges of generalization in this domain and motivating future research in metadata extraction from municipal minutes.

Abstract
State-based Conflict-Free Replicated Data Types (CRDTs) are widely used in distributed systems to ensure high availability without coordination. However, their naive synchronization strategy, transmitting the full state, incurs high communication costs. In this paper, we: (1) propose ConflictSync, a digest-driven synchronization algorithm, which reduces total data transfer by up to 18× compared to full-state transmissions; (2) formulate state-based CRDT synchronization as set reconciliation over irredundant join decompositions; (3) generalize Rateless Set Reconciliation for variable-sized elements, at the cost of an additional communication step; (4) introduce a new generic set reconciliation solution, integrating Bloom Filters with rateless IBLTs; (5) experimentally evaluate the novel synchronization strategies.

Abstract
This paper presents a novel hybrid deep learning framework designed to enhance the robustness of CSI-based Human Activity Recognition (HAR) within bandwidth-constrained Wi-Fi sensing environments. The core of our proposed methodology is a preliminary Doppler trace extraction stage, implemented to amplify salient motion-related signal features before classification. Subsequently, these enhanced inputs are processed by a hybrid neural architecture, which integrates Inception networks responsible for hierarchical spatial feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) networks that capture temporal dependencies. A Support Vector Machine (SVM) is then utilized as the final classification layer to optimize decision boundaries. The framework's efficacy was systematically validated using a public dataset across 20, 40, and 80 MHz bandwidth configurations. The model yielded accuracies of 89.27%(20 MHz), 94.13%(40 MHz), and 95.30%(80 MHz), respectively. These results confirm a marked superiority over standalone deep learning baselines, especially in the most constrained low-bandwidth scenarios. This study underscores the utility of combining Doppler-based feature engineering with a hybrid learning architecture for reliable HAR in bandwidth-limited wireless sensing applications. © 2026 IEEE.

Abstract
Conflict-free Replicated Data Types (CRDTs) enable available and eventually consistent data replication without coordination, making them well suited for open and partition-prone environments. Recent work has shown that CRDTs can be extended to tolerate Byzantine faults by ensuring that replicas eventually agree on the validity of operations, even in permis-sionless settings. However, validity alone does not prevent a Byzantine participant from inflicting unbounded damage by issuing large volumes of adversarial yet well-formed updates. For example, when editing text, an attacker can easily delete prior text. In this paper, we study how to bound the impact of Byzantine behavior in open CRDT systems. We introduce bounded Byzantine CRDTs, a rate-limiting framework for CRDTs in which each update carries an associated cost that limits the influence of adversarial operations relative to the resources they expend. Overall, this work bridges the gap between Byzantine-Tolerant CRDTs and resource-bounded adversarial models, providing a principled foundation for deploying CRDTs in fully open, adversarial environments.

Abstract
Unmanned Aerial Vehicles (UAVs) have emerged as key enablers in Non-Terrestrial Networks (NTNs) to provide flexible wireless coverage, particularly in infrastructure-limited scenarios. In our previous work, we proposed the Sustainable multi-UAV Performance-aware Placement (SUPPLY) algorithm, a pioneering solution for the energy-efficient placement of multiple UAVs acting as Flying Access Points (FAPs). SUPPLY ensures continuous Ground User (GU) coverage while minimizing propulsion energy consumption. However, its quadratic time complexity in the GU grouping phase imposes scalability constraints, especially in larger time-sensitive scenarios. In this paper, we propose eSUPPLY, a computationally efficient enhancement to SUPPLY. By increasing the step size between candidate Flying Access Point (FAP) positions during the GU grouping phase, eSUPPLY significantly reduces the size of the optimization problem. Simulation results demonstrate up to a 97% reduction in execution time, with only a marginal increase in the number of FAPs and energy consumption, while maintaining network performance, enabling real-time operation in larger and more dynamic Flying Networks (FNs) compared to the SUPPLY algorithm.

Abstract
The link between epicardial adipose tissue (EAT) and cardiovascular risk is well established, with EAT volume being strongly associated with inflammation, coronary artery disease (CAD) risk, and mortality. However, its EAT quantification is hindered by the time-consuming nature of manual EAT segmentation in cardiac computed tomography (CT). 300 non-contrast cardiac CT scans were collected and the pericardium was manually delineated. In a subset of this data (N = 30), manual delineation was repeated by the same operator and by a second operator. Two automatic methods were then used for pericardial segmentation: a commercially available tool, Siemens Cardiac Risk Assessment (CRA) software; and a deep learning solution based on a U-Net architecture trained exclusively with external public datasets (CardiacFat and OSIC). EAT segmentations were obtained through thresholding to [- 150,- 50] Hounsfield units. Pericardial and EAT segmentation performance was evaluated considering the segmentations by the first operator as reference. Statistical significance of differences for all metrics and segmentation methods was tested through Student t-tests. Pericardial segmentation intra-/interobserver variability was excellent, with the U-Net outperforming Siemens CRA (p < 0.0001). The intra- and interobserver agreement for EAT segmentation was lower with Dice Scores (DSC) of 0.862 and 0.775 respectively, while the U-Net and Siemens CRA obtained DSCs of 0.723 and 0.679 respectively. EAT volume quantification showed that the agreement between a human observer and the U-Net was better than that of two human observers (p = 0.0141), with a Pearson Correlation Coefficient (PCC) of 0.896 and a bias of - 2.83 cm(3) (below the interobserver bias of 9.05 cm3). The lower performances of EAT segmentation highlight the difficulty in segmenting this structure. For both pericardial and EAT segmentation, the deep learning method outperformed the commercial solution. While the segmentation performance of the U-Net solution was below interobserver variability, EAT volume quantification performance was competitive with human readers, motivating future use of these tools. Clinical trial number: NCT03280433, registered retrospectively on 2017-09-08.