Abstract
This study proposes a comprehensive framework integrating eXplainable Artificial Intelligence (XAI) techniques with clustering-based context extraction to enhance energy consumption forecasting in modern office buildings. By leveraging explanation vectors derived from state-of-the-art XAI methods such as SHAP and LIME, our framework identifies latent operational contexts from sensor data aggregated at 15-min intervals. These contexts enable the tailoring of predictive models through feature augmentation, context-specific training, and transfer learning strategies, thereby improving forecasting accuracy compared to conventional approaches. To identify the best-performing models for each context, hyperparameter optimization via grid search is employed across multiple algorithmsincluding Gradient Boosting, Random Forest, and K-Nearest Neighbors. Extensive experiments demonstrate that context-aware models significantly outperform baseline methods, achieving up to a 7% improvement in the coefficient of determination (R-2) and a marked reduction in error metrics. Our findings underscore the importance of integrating XAI with data-driven modeling to enhance predictive performance and model interpretability, which are critical for practical energy management and decision-making in complex building environments.

Abstract
The production of insects for both human and animal consumption has seen an interest increase in recent years. Tenebrio molitor, a beetle species whose larval stage was considered by the European Commission safe for human consumption, represents a promising candidate for large-scale production. Efficient production of T. molitor requires the separation of its different metamorphic stages to optimize productivity and reduce cannibalism. Additionally, dead pupae must be removed to prevent contamination and ensure the healthy development of the remaining individuals. Currently, the identification of dead pupae relies primarily on visual inspection based on colour changes, as dead individuals tend to exhibit surface melanization. However, this method becomes inefficient in large-scale operations. This study presents a benchmark of hyperspectral imaging (HSI) and thermal imaging as alternative sensing technologies for automated dead pupae identification. Hyperspectral data acquired in the near-infrared range (900-1700 nm) enabled accurate discrimination between dead and live pupae using both a PLS-DA and a Logistic Regression model, achieving F1-Scores above 90 % for both classes. Furthermore, key wavelengths (903 nm and 1259 nm) were identified and used to develop a normalized difference index (NDI) capable of distinguishing pupae health status using only two spectral bands. Thermal imaging revealed consistent temperature differences, with dead pupae presenting approximately 1-3 % lower temperatures than live pupae. The proven reliability and precision of the proposed sensing technologies validate their use in contributing to the development of scalable and reliable solutions for quality control in the insect farming industry.

Abstract

Abstract

Abstract
Accurate segmentation of grapevines in imagery acquired from unmanned aerial vehicles (UAVs) is important for precision viticulture, as it supports vineyard management by monitoring grapevine health, growth, and environmental stress. However, the structural diversity of vineyards, including differences in training systems, row curvatures, and foliage density, presents challenges for grapevine segmentation methods. This study evaluates the performance of deep learning (DL) models-Feature Pyramid Network (FPN), Pyramid Scene Parsing Network (PSPNet) and U-Net-each combined with different backbones for grapevine segmentation in UAV-based RGB orthophoto mosaics. Data were collected under a range of vineyard conditions and scenarios from Portugal's Douro and Vinhos Verdes regions, providing a representative dataset across multiple vineyard configurations. The DL models were trained, tested, and evaluated using orthorectified RGB imagery, and their segmentation accuracy was compared to thresholding techniques. The results show that DL models, particularly U-Net, achieved accurate grapevine segmentation and reduced over-segmentation and false detections that are common in thresholding methods. FPN models with Inception-v4 and Xception backbones performed well in vineyards with inter-row vegetation, while PSPNet models showed segmentation limitations. Overall, DL-based segmentation models demonstrated advantages over thresholding approaches, demonstrating their suitability for UAV-based grapevine segmentation in diverse and challenging vineyard environments. These results support the scalability of DL-based segmentation for vineyard monitoring applications and indicate that improved segmentation accuracy can contribute to decision support in precision viticulture.

Abstract
Validation is a central activity when developing formal specifications. Similarly to coding, a possible validation technique is to define upfront test cases or scenarios that a future specification should satisfy or not. Unfortunately, specifying such test cases is burdensome and error prone, which could cause users to skip this validation task. This paper reports the results of an empirical evaluation of using pre-trained large language models (LLMs) to automate the generation of test cases from natural language requirements. In particular, we focus on test cases for structural requirements of simple domain models formalized in the Alloy specification language. Our evaluation focuses on the state-of-theart GPT-5 model, but results from other closed- and open-source LLMs are also reported. The results show that, in this context, GPT-5 is already quite effective at generating positive (and negative) test cases that are syntactically correct and that satisfy (or not) the given requirement, and that can detect many wrong specifications written by humans.