Argument mining (AM) is an interdisciplinary research field focused on the automatic identification and classification of argumentative components, such as claims and premises, and the relationships between them. Recent advances in large language models (LLMs) have significantly improved the performance of argument classification compared to traditional machine learning approaches. However, there remains a lack of systematic comparative evaluation of modern LLMs across multiple benchmark datasets, as well as a limited understanding of their error patterns and failure modes. This study presents a comprehensive evaluation of several state-of-the-art LLMs, including GPT-5.2, Llama 4, and DeepSeek R1, on large publicly available argument classification corpora such as Args.me and UKP. The evaluation incorporates advanced prompting strategies, including Chain-of-Thought prompting, prompt rephrasing, voting, and certainty-based classification. Both quantitative performance metrics and qualitative error analysis are conducted to assess model behavior. The best-performing model in the study (GPT-5.2) achieves a classification accuracy of 78.0

Cross-entropy loss has long been the standard choice for training deep neural networks, yet it suffers from interpretability limitations, unbounded weight growth, and inefficiencies that can contribute to costly training dynamics. The harmonic loss is a distance-based alternative grounded in Euclidean geometry that improves interpretability and mitigates phenomena such as grokking, or delayed generalization on the test set. However, the study of harmonic loss remains narrow: only Euclidean distance is explored, and no systematic evaluation of computational efficiency or sustainability was conducted. We extend harmonic loss by systematically investigating a broad spectrum of distance metrics as replacements for the Euclidean distance. We comprehensively evaluate distance-tailored harmonic losses on both vision backbones and large language models. Our analysis is framed around a three-way evaluation of model performance, interpretability, and sustainability. On vision tasks, cosine distances provide the most favorable trade-off, consistently improving accuracy while lowering carbon emissions, whereas Bray-Curtis and Mahalanobis further enhance interpretability at varying efficiency costs. On language models, cosine-based harmonic losses improve gradient and learning stability, strengthen representation structure, and reduce emissions relative to cross-entropy and Euclidean heads. Our code is available at: https://anonymous.4open.science/r/rethinking-harmonic-loss-5BAB/.

Salient Object Detection in Optical Remote Sensing Images (ORSI-SOD) is vital for applications such as urban planning and disaster monitoring. Yet, existing Artificial Neural Network (ANN) models remain energy-intensive and unsuitable for edge deployment. Based on this, we propose a Low-ENergy Spiking Neural Network (LENS-Net) for remote sensing saliency, the first fully spiking neural network for ORSI-SOD. LENS-Net employs an improved Spike-Driven Transformer v3 (SDTv3) encoder to extract multi-scale features with low-energy cost, and a novel Spike Multi-scale Attention Decoder (SpikeMAD) that fuses contextual cues via spike-driven channel attention and up-convolution, ensuring effective saliency representation under sparse computation. Moreover, to stabilize training and improve boundary recognition in complex scenes, we propose a Soft-Clip Spike Firing Approximation (SC-SFA). Across three benchmark datasets, LENS-Net demonstrates outstanding performance while maintaining high energy efficiency, outperforming all lightweight ANN counterparts. For instance, on the ORSSD dataset under a timestep T =4 during inference, LENS-Net achieves a S-alpha of 92.79%, an MAE of 0.0109, and consumes only 11.48 mJ of energy. These results establish an efficient, low-energy solution for practical ORSI-SOD deployment. The source code is available at: https://github.com/7dra/LENS-Net.

This article concerns comparative studies on the Automatic Speech Recognition (ASR) model incorporated with the Large Language Model (LLM) used for medical interviews. The proposed solution is tested on polish language benchmarks and dataset with medical interviews. The latest ASR technologies are based on convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers. Most of them work as end-to-end solutions. The presented approach in the case of the Whisper model shows a two-stage solution with End-To-End ASR and LLM working together in a pipeline. The ASR output is an input for LLM. The LLM is a component by which the output from ASR is corrected and improved. Comparative studies for automatic recognition of the Polish language between modern End-To-End deep learning architectures and the ASR hybrid model were performed. The medical interview tests were performed with two state-of-the-art ASR models: OpenAI Whisper incorporated with LLM and Scribe ElevenLabs. Additionally, the results were compared with five more end-to-end models (QuartzNet, FastConformer, Wav2Vec 2.0 XLSR and ESPnet Model Zoo) on Mozilla Common Voice and VoxPopuli databases. Tests were conducted for clean audio signal, signal with bandwidth limitation, and degraded. The tested models were evaluated on the basis of Word Error Rate (WER) and Character Error Rate (CER). The results show that the Whisper model performs by far the best among the open-source models. ElevenLabs Scribe model, on the other hand, performs best for Polish on both general benchmark and medical data.

Continual offline reinforcement learning (CORL) aims to learn a sequence of tasks from datasets collected over time while preserving performance on previously learned tasks. This setting corresponds to domains where new tasks arise over time, but adapting the model in live environment interactions is expensive, risky, or impossible. However, CORL inherits the dual difficulty of offline reinforcement learning and adapting while preventing catastrophic forgetting. Replay-based continual learning approaches remain a strong baseline but incur memory overhead and suffer from a distribution mismatch between replayed samples and newly learned policies. At the same time, architectural continual learning methods have shown strong potential in supervised learning but remain underexplored in CORL. In this work, we propose TSN-Affinity, a novel CORL method based on TinySubNetworks and Decision Transformer. The method enables task-specific parameterization and controlled knowledge sharing through a RL-aware reuse strategy that routes tasks according to action compatibility and latent similarity. We evaluate the approach on benchmarks based on Atari games and simulations of manipulation tasks with the Franka Emika Panda robotic arm, covering both discrete and continuous control. Results show strong retention from sparse SubNetworks, with routing further improving multi-task performance. Our findings suggest that similarity-guided architectural reuse is a strong and viable alternative to replay-based strategies in a CORL setting. Our code is available at: https://github.com/anonymized-for-submission123/tsn-affinity.

Reinforcement learning (RL) in robotics faces significant hurdles regarding sample efficiency and generalization across varying goals. While Offline RL mitigates the need for costly online interactions, its integration with goal-conditioned policies and transformer-based architectures remains underexplored. We introduce a Goal-Conditioned Decision Transformer adapted for offline multi-goal robotics. By explicitly incorporating goal states into the sequence modeling framework, our approach efficiently solves varying tasks using only pre-collected data. We validate this method on a newly released offline dataset for the Franka Emika Panda platform. Experimental results demonstrate that our approach outperforms state-of-the-art online baselines in complex tasks and maintains robustness in sparse-reward settings, even with limited expert demonstrations.

Deep learning models are the most efficient models in many machine learning tasks. The main disadvantage when using them in IoT, mobile devices, independent autonomous or real-time systems is their complexity and memory size. Therefore, much research has concentrated on compression techniques of deep learning architectures. One of the most popular technique is quantization. In most of the works, the quantization is done based on the nearest neighbour quantization technique. This work focuses on improving the quantization efficiency in pretrained and quantized models. This approach has the potential to improve the final accuracy of quantized models. The main postulate of the work is that final quantization states of the network based on nearest neighbour rounding does not guarantee optimal accuracy. In the presented work, the evolution strategy is used as an optimization approach. The evolution in each iteration changes the values of the small percentage of weights. It shifts theirs values to different quantization states. The work shows that proposed evolution with an appropriate set of operators and parameters can fast improve the accuracy of the quantized models. The results are presented for popular architectures such as VGG and Resnet for image classification and detection. Additionally, simulations were carried out for the autoencoder architecture.

Anomaly detection tools and methods present a key capability in modern cyberphysical and failure prediction systems. Despite the fast-paced development in deep learning architectures for anomaly detection, model optimization for a given dataset is a cumbersome and time-consuming process. Neuroevolution could be an effective and efficient solution to this problem, as a fully automated search method for learning optimal neural networks, supporting both gradient and nongradient fine-tuning. However, existing methods mostly focus on optimizing model architectures without taking into account feature subspaces and model weights. In this work, we propose anomaly detection neuroevolution (AD-NEv)-a scalable multilevel optimized neuroevolution framework for multivariate time-series anomaly detection. The method represents a novel approach to synergically: 1) optimize feature subspaces for an ensemble model based on the bagging technique; 2) optimize the model architecture of single anomaly detection models; and 3) perform nongradient fine-tuning of network weights. An extensive experimental evaluation on widely adopted multivariate anomaly detection benchmark datasets shows that the models extracted by AD-NEv outperform well-known deep learning architectures for anomaly detection. Moreover, results show that AD-NEv can perform the whole process efficiently, presenting high scalability when multiple graphics processing units (GPUs) are available.

Post-hoc explainability methods typically associate each output score of a deep neural network with an input-space direction, most commonly instantiated as the gradient and visualized as a saliency map. However, these approaches often yield explanations that are noisy, lack perceptual alignment and, thus, offer limited interpretability. While many explanation methods attempt to address this issue via modified backward rules or additional heuristics, such approaches are often difficult to justify theoretically and frequently fail basic sanity checks. We introduce Semantic Pullbacks (SP), a faithful and effective post-hoc explanation method for deep neural networks. Semantic Pullbacks address the limitations above by isolating the network's effective linear action via a principled pullback formulation and refining it to recover coherent local structures learned by the target neuron. As a result, SP produces perceptually aligned, class-conditional explanations that highlight meaningful features, support compelling counterfactual perturbations, and admit a clear theoretical motivation. Across standard faithfulness benchmarks, Semantic Pullbacks significantly outperform established attribution methods on both classical convolutional architectures (ResNet50, VGG) and transformer-based models (PVT), while remaining general and computationally efficient. Our method can be easily plugged into existing deep learning pipelines and extended to other modalities.

Flight schedules are an essential part of airline operations and impact all key elements, such as airplanes, crew, and passengers. Therefore, efficient flight scheduling is crucial for optimizing airline resource utilization and minimizing operational costs. Unfortunately, unexpected delays and interruptions can significantly affect even the most optimized flight schedules. Therefore, airlines must be able to recover efficiently from such situations to protect their revenue and reputation. In this context, artificial intelligence (AI) methods can be used to identify suitable recovery solutions when unexpected events occur. The methods proposed so far often leverage meta-heuristic methods, conventional optimization, and, more recently, machine learning. However, current works neither leverage integer linear programming (ILP) nor focus on simple scenarios that do not fully entail real-world complexities. This creates a significant gap, as ILP has the potential to provide effective rescheduling capabilities. To this end, we propose a novel approach to aircraft recovery problems based on integer linear programming. We formulate efficient equations that correspond to the restrictions, objective functions, and possible recovery actions. The model aims to minimize the total delays caused by disruptions by swapping aircraft and delaying flights as recovery options. Our experimental study involving two real-world airline datasets shows that the proposed method is effective in providing a significant reduction in delays while maintaining a limited execution time.

Argument mining (AM) is an interdisciplinary research field that integrates insights from logic, philosophy, linguistics, rhetoric, law, psychology, and computer science. It involves the automatic identification and extraction of argumentative components, such as premises and claims, and the detection of relationships between them, such as support, attack, or neutrality. Recently, the field has advanced significantly, especially with the advent of large language models (LLMs), which have enhanced the efficiency of analyzing and extracting argument semantics compared to traditional methods and other deep learning models. There are many benchmarks for testing and verifying the quality of LLM, but there is still a lack of research and results on the operation of these models in publicly available argument classification databases. This paper presents a study of a selection of LLM's, using diverse datasets such as Args.me and UKP. The models tested include versions of GPT, Llama, and DeepSeek, along with reasoning-enhanced variants incorporating the Chain-of-Thoughts algorithm. The results indicate that ChatGPT-4o outperforms the others in the argument classification benchmarks. In case of models incorporated with reasoning capabilities, the Deepseek-R1 shows its superiority. However, despite their superiority, GPT-4o and Deepseek-R1 still make errors. The most common errors are discussed for all models. To our knowledge, the presented work is the first broader analysis of the mentioned datasets using LLM and prompt algorithms. The work also shows some weaknesses of known prompt algorithms in argument analysis, while indicating directions for their improvement. The added value of the work is the in-depth analysis of the available argument datasets and the demonstration of their shortcomings.

This paper contributes to solving the low auto correlation binary sequence (LABS) problem that remains an open hard-optimization problem with many applications. The current direction of research is focused on developing algorithms dedicated to parallel architectures such as GPGPU or multi-core CPUs. The paper follows this direction and proposes new heuristics developed from the steepest-descent local search algorithm that extends the notion of a neighborhood of a given sequence. The introduced algorithms utilize the parallel nature of multicore CPUs and provide an effective method for solving the LABS problem. The efficiency levels of SDSL and the new algorithm are presented; to ensure an effective comparison, they were both implemented in the same manner. The comparison shows that exploring the larger neighborhood improves the efficiency of the search method.

The recently proposed xLSTM is a powerful model that leverages expressive multiplicative gating and residual connections, providing the temporal capacity needed for long-horizon forecasting and representation learning. This architecture has demonstrated success in time series forecasting, lossless compression, and even large-scale language modeling tasks, where its linear memory footprint and fast inference make it a viable alternative to Transformers. Despite its growing popularity, no prior work has explored xLSTM for anomaly detection. In this work, we fill this gap by proposing xLSTMAD, the first anomaly detection method that integrates a full encoder-decoder xLSTM architecture, purpose-built for multivariate time series data. Our encoder processes input sequences to capture historical context, while the decoder is devised in two separate variants of the method. In the forecasting approach, the decoder iteratively generates forecasted future values xLSTMAD-F, while the reconstruction approach reconstructs the input time series from its encoded counterpart xLSTMAD-R. We investigate the performance of two loss functions: Mean Squared Error (MSE), and Soft Dynamic Time Warping (SoftDTW) to consider local reconstruction fidelity and global sequence alignment, respectively. We evaluate our method on the comprehensive TSB-AD-M benchmark, which spans 17 real-world datasets, using state-of-the-art challenging metrics such as VUS-PR. In our results, xLSTM showcases state-of-the-art accuracy, outperforming 23 popular anomaly detection baselines. Our paper is the first work revealing the powerful modeling capabilities of xLSTM for anomaly detection, paving the way for exciting new developments on this subject. Our code is available at: https://github.com/Nyderx/xlstmad

Multivariate time series anomaly detection is a crucial problem in many industrial and research applications. Timely detection of anomalies allows, for instance, to prevent defects in manufacturing processes and failures in cyberphysical systems. Deep learning methods are preferred among others for their accuracy and robustness for the analysis of complex multivariate data. However, a key aspect is being able to extract predictions in a timely manner, to accommodate real-time requirements in different applications. In the case of deep learning models, model reduction is extremely important to achieve optimal results in real-time systems with limited time and memory constraints. In this paper, we address this issue by proposing a novel compression method for deep autoencoders that involves three key factors. First, pruning reduces the number of weights, while preventing catastrophic drops in accuracy by means of a fast search process that identifies high sparsity levels. Second, linear and non-linear quantization reduces model complexity by reducing the number of bits for every single weight. The combined contribution of these three aspects allow the model size to be reduced, by removing a subset of the weights (pruning), and decreasing their bit-width (quantization). As a result, the compressed model is faster and easier to adopt in highly constrained hardware environments. Experiments performed on popular multivariate anomaly detection benchmarks, show that our method is capable of achieving significant model compression ratio (between 80% and 95%) without a significant reduction in the anomaly detection performance.

Reinforcement learning (RL) in robotics faces significant hurdles regarding sample efficiency and generalization across varying goals. While Offline RL mitigates the need for costly online interactions, its integration with goal-conditioned policies and transformer-based architectures remains underexplored. We introduce a Goal-Conditioned Decision Transformer adapted for offline multi-goal robotics. By explicitly incorporating goal states into the sequence modeling framework, our approach efficiently solves varying tasks using only pre-collected data. We validate this method on a newly released offline dataset for the Franka Emika Panda platform. Experimental results demonstrate that our approach outperforms state-of-the-art online baselines in complex tasks and maintains robustness in sparse-reward settings, even with limited expert demonstrations.

Continual learning (CL) is one of the most promising trends in recent machine learning research. Its goal is to go beyond classical assumptions in machine learning and develop models and learning strategies that present high robustness in dynamic environments. This goal is realized by designing strategies that simultaneously foster the incorporation of new knowledge while avoiding forgetting past knowledge. The landscape of CL research is fragmented into several learning evaluation protocols, comprising different learning tasks, datasets, and evaluation metrics. Additionally, the benchmarks adopted so far are still distant from the complexity of real-world scenarios, and are usually tailored to highlight capabilities specific to certain strategies. In such a landscape, it is hard to clearly and objectively assess models and strategies. In this work, we fill this gap for CL on image data by introducing two novel CL benchmarks that involve multiple heterogeneous tasks from six image datasets, with varying levels of complexity and quality. Our aim is to fairly evaluate current state-of-the-art CL strategies on a common ground that is closer to complex real-world scenarios. We additionally structure our benchmarks so that tasks are presented in increasing and decreasing order of complexity—according to a curriculum—in order to evaluate if current CL models are able to exploit structure across tasks. We devote particular emphasis to providing the CL community with a rigorous and reproducible evaluation protocol for measuring the ability of a model to generalize and not to forget while learning. Furthermore, we provide an extensive experimental evaluation showing that popular CL strategies, when challenged with our proposed benchmarks, yield sub-par performance, high levels of forgetting, and present a limited ability to effectively leverage curriculum task ordering. We believe that these results highlight the need for rigorous comparisons in future CL works as well as pave the way to design new CL strategies that are able to deal with more complex scenarios.

Unmanned aerial vehicle (UAV) photogrammetry allows the generation of orthophoto and digital surface model (DSM) rasters of terrain. However, DSMs of water bodies mapped using this technique often reveal distortions in the water surface, thereby impeding the accurate sampling of water surface elevation (WSE) from DSMs. This study investigates the capability of deep neural networks to accommodate the aforementioned perturbations and effectively estimate WSE from photogrammetric rasters. Convolutional neural networks (CNNs) were employed for this purpose. Two regression approaches utilizing CNNs were explored: direct regression employing an encoder and a solution based on prediction of the weight mask by an autoencoder architecture, subsequently used to sample values from the photogrammetric DSM. The dataset employed in this study comprises data collected from five case studies of small lowland streams in Poland and Denmark, consisting of 322 DSM and orthophoto raster samples. A grid search was employed to identify the optimal combination of encoder, mask generation architecture, and batch size among multiple candidates. Solutions were evaluated using two cross-validation methods: stratified k-fold cross-validation, where validation subsets maintained the same proportion of samples from all case studies, and leave-one-case-out cross-validation, where the validation dataset originates entirely from a single case study, and the training set consists of samples from other case studies. Depending on the case study and the level of validation strictness, the proposed solution achieved a root mean square error (RMSE) ranging between 2 cm and 16 cm. The proposed method outperforms methods based on the straightforward sampling of photogrammetric DSM, achieving, on average, an 84% lower RMSE for stratified cross-validation and a 62% lower RMSE for all-in-case-out cross-validation. By utilizing data from other research, the proposed solution was compared on the same case study with other UAV-based methods. For that benchmark case study, the proposed solution achieved an RMSE score of 5.9 cm for all-in-case-out cross-validation and 3.5 cm for stratified cross-validation, which is close to the result achieved by the radar-based method (RMSE of 3 cm), which is considered the most accurate method available. The proposed solution is characterized by a high degree of explainability and generalization.

Efficient flight scheduling is crucial to properly allocate airline resources, but even the best flight schedule has to face unexpected delays and disruptions. The ability to recover from such disruptions is essential for airlines to minimize the negative impact on their revenue and reputation. In this context, machine learning-based methods can be used to identify suitable recovery methods as unexpected events occur. Reinforcement learning approaches are especially promising since they extract suitable solutions much more efficiently than conventional optimization and meta-heuristics methods and provide timely rescheduling capabilities for airlines, which translates into reduced capital and reputation losses. However, current works either do not leverage deep learning or focus on simple scenarios that do not fully entail real-world complexities, resulting in limited efficiency or sub-optimal solutions. In this paper, we propose an ensemble of two deep learning approaches: Deep Double Q-Learning (DDQL) and Advantage Actor-Critic (A2C). The models aim to minimize the total delays caused by disruptions by swapping aircraft and delaying flights as recovery options. We perform experiments with a benchmark dataset and a real-world airline dataset, showing that our method is effective in providing a significant reduction of delays caused by disruptions.

Anomaly detection tools and methods enable key analytical capabilities in modern cyberphysical and sensor-based systems. Despite the fast-paced development in deep learning architectures for anomaly detection, model optimization for a given dataset is a cumbersome and time-consuming process. Neuroevolution could be an effective and efficient solution to this problem, as a fully automated search method for learning optimal neural networks, supporting both gradient and non-gradient fine tuning. However, existing frameworks incorporating neuroevolution lack of support for new layers and architectures and are typically limited to convolutional and LSTM layers. In this paper we propose AD-NEv++, a three-stage neuroevolution-based method that synergically combines subspace evolution, model evolution, and fine-tuning. Our method overcomes the limitations of existing approaches by optimizing the mutation operator in the neuroevolution process, while supporting a wide spectrum of neural layers, including attention, dense, and graph convolutional layers. Our extensive experimental evaluation was conducted with widely adopted multivariate anomaly detection benchmark datasets, and showed that the models generated by AD-NEv++ outperform well-known deep learning architectures and neuroevolution-based approaches for anomaly detection. Moreover, results show that AD-NEv++ can improve and outperform the state-of-the-art GNN (Graph Neural Networks) model architecture in all anomaly detection benchmarks.