Abstract
The interaction between humans and technology has always been a key determinant factor of adoption and efficiency. This is true whether the interaction is with hardware, software or data. In the particular case of Information Retrieval (IR), recent developments in Deep Learning and Natural Language Processing (NLP) techniques opened the door to more natural and efficient IR means, no longer based on keywords or similarity metrics but on a distributed representation of meaning. In this paper we propose an agent-based architecture to serve as an interface with industrial systems, in which agents are powered by specific Large Language Models (LLMs). Its main goal is to make the interaction with such systems (e.g. data sources, production systems, machines) natural, allowing users to execute complex tasks with simple prompts. To this end, key aspects considered in the architecture are human-centricity and context-awareness. This paper provides a high-level description of this architecture, and then focuses on the development and evaluation of one of its key agents, responsible for information retrieval. For this purpose, we detail three application scenarios, and evaluate the ability of this agent to select the appropriate data sources to answer a specific prompt. Depending on the scenario and on the underlying model, results show an accuracy of up to 80%, showing that the proposed agent can be used to autonomously select from among several available data sources to answer a specific information need. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
The effectiveness of time series forecasting models can be hampered by conditions in the input space that lead them to underperform. When those are met, negative behaviours, such as higher-than-usual errors or increased uncertainty are shown. Traditionally, stress testing is applied to assess how models respond to adverse, but plausible scenarios, providing insights on how to improve their robustness and reliability. This paper builds upon this technique by contributing with a novel framework called MAST (Meta-learning and data Augmentation for Stress Testing). In particular, MAST is a meta-learning approach that predicts the probability that a given model will perform poorly on a given time series based on a set of statistical features. This way, instead of designing new stress scenarios, this method uses the information provided by instances that led to decreases in forecasting performance. An additional contribution is made, a novel time series data augmentation technique based on oversampling, that improves the information about stress factors in the input space, which elevates the classification capabilities of the method. We conducted experiments using 6 benchmark datasets containing a total of 97.829 time series. The results suggest that MAST is able to identify conditions that lead to large errors effectively. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
Multi-robot coordination aims to synchronize robots for optimized, collision-free paths in shared environments, addressing task allocation, collision avoidance, and path planning challenges. The Time Enhanced A* (TEA*) algorithm addresses multi-robot pathfinding offering a centralized and sequential approach. However, its sequential nature can lead to order-dependent variability in solutions. This study enhances TEA* through multi-threading, using thread pooling and parallelization techniques via OpenMP, and a sensitivity analysis enabling parallel exploration of robot-solving orders to improve robustness and the likelihood of finding efficient, feasible paths in complex environments. The results show that this approach improved coordination efficiency, reducing replanning needs and simulation time. Additionally, the sensitivity analysis assesses TEA*'s scalability across various graph sizes and number of robots, providing insights into how these factors influence the efficiency and performance of the algorithm. © 2025 IEEE.

Abstract
Accurate evaluation of forecasting models is essential for ensuring reliable predictions. Current practices for evaluating and comparing forecasting models focus on summarising performance into a single score, using metrics such as SMAPE. We hypothesize that averaging performance over all samples dilutes relevant information about the relative performance of models. Particularly, conditions in which this relative performance is different than the overall accuracy. We address this limitation by proposing a novel framework for evaluating univariate time series forecasting models from multiple perspectives, such as one-step ahead forecasting versus multi-step ahead forecasting. We show the advantages of this framework by comparing a state-of-the-art deep learning approach with classical forecasting techniques. While classical methods (e.g. ARIMA) are long-standing approaches to forecasting, deep neural networks (e.g. NHITS) have recently shown state-of-the-art forecasting performance in benchmark datasets. We conducted extensive experiments that show NHITS generally performs best, but its superiority varies with forecasting conditions. For instance, concerning the forecasting horizon, NHITS only outperforms classical approaches for multi-step ahead forecasting. Another relevant insight is that, when dealing with anomalies, NHITS is outperformed by methods such as Theta. These findings highlight the importance of evaluating forecasts from multiple dimensions.

Abstract
The detection of Pulmonary Hypertension (PH) from the computer analysis of digitized heart sounds is a low-cost and non-invasive solution for early PH detection and screening. We present an extensive cross-domain evaluation methodology with varying animals (humans and porcine animals) and varying auscultation technologies (phonocardiography and seisomocardiography) evaluated across four methods. We introduce PH-ELM, a resource-efficient PH detection model based on the extreme learning machine that is smaller (300× fewer parameters), energy efficient (532× fewer watts of power), faster (36× faster to train, 44× faster at inference), and more accurate on out-of-distribution testing (improves median accuracy by 0.09 area under the ROC curve (auROC)) in comparison to a previously best performing deep network. We make four observations from our analysis: (a) digital auscultation is a promising technology for the detection of pulmonary hypertension; (b) seismocardiography (SCG) signals and phonocardiography (PCG) signals are interchangeable to train PH detectors; (c) porcine heart sounds in the training data can be used to evaluate PH from human heart sounds (the PH-ELM model preserves 88 to 95% of the best in-distribution baseline performance); (d) predictive performance of PH detection can be mostly preserved with as few as 10 heartbeats and capturing up to approximately 200 heartbeats per subject can improve performance. © 1964-2012 IEEE.

Abstract
The International Symposium on Applied Reconfigurable Computing (ARC) is an annual forum for the discussion and dissemination of research, notably applying the Reconfigurable Computing (RC) concept to real-world problems. The first edition of ARC took place in 2005, and in 2024, ARC celebrated its 20th edition. During those 20 years, the field of reconfigurable computing saw a tremendous growth in its underlying technology. ARC contributed very significantly to the presentation and dissemination of new ideas, innovative applications, and fruitful discussions, all of which have resulted in the shaping of novel lines of research. Here, we present selected papers from the first 20 years of ARC, that we believe represent the corpus of work and reflect the ARC spirit by covering a broad spectrum of RC applications, benchmarks, tools, and architectures. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.