Continual anomaly detection in tabular data is challenging and remains largely underexplored, particularly in settings with heterogeneous feature schemas, distribution shifts, and severe class imbalance. In many real-world applications, data arrive sequentially from diverse domains, rendering conventional continual learning methods ineffective due to their reliance on a fixed input space. We propose a continual learning (CL) method, which can overcome these challenges and continually learn from different tasks. Our method consists of three main parts: our AGF model, Taskfusion augmentation, and outlier exposure. The AGF-model maps task-specific features into a shared space, then aligns distributions to reduce representation drift, and learns anomaly decision boundaries in the aligned space. To improve stability, we introduce Taskfusion augmentation, combining boundary-aware interpolation within tasks to refine the model anomaly boundaries and cross-task mixing to transfer anomaly structure across datasets. To handle class imbalance and memory constraints, we employ tabular dataset distillation to store compact synthetic replay samples, which are jointly used with augmented data in an outlier exposure objective for robust anomaly detection. We evaluate the approach on 21 heterogeneous datasets across multiple domains. Results show that our approach substantially improves continual anomaly detection performance over sequential fine-tuning and other CL baselines while reducing catastrophic forgetting and maintaining stable detection across heterogeneous datasets.

Very-high-resolution remote-sensing imagery provides a scalable basis for delineating informal settlements, but sparse annotations, severe imbalance between informal-settlement and background pixels, and cross-city heterogeneity in urban morphology and image–mask correspondence complicate model development. We present SLUM-i, a semi-supervised semantic segmentation framework together with a geographically diverse seven-city Earth observation benchmark spanning three continents. The benchmark combines a newly annotated Lahore dataset and companion Karachi and Mumbai datasets with four publicly released city datasets, totaling 14,458 RGB image–mask tiles. We quantify cross-city heterogeneity using class composition, boundary morphology, grayscale separability, correspondence between mask boundaries and image edges, and divergence between measured feature distributions. For label-efficient mapping, SLUM-i combines representation-guided unlabeled-pool curation, using embeddings from a vision foundation model (DINOv2-Small) to remove the least-similar tiles, and Class-Aware Adaptive Thresholding, which adapts pseudo-label acceptance by class through global mean-confidence and per-class mean-softmax exponential moving averages. Experiments at 10

Histology-based single-cell spatial transcriptomics (ST) estimation aims to predict gene expression for individual cells from histopathological images and cell locations, reducing the need for costly single-cell ST measurements. Unlike existing histology-to-ST methods that mainly predict spot-level profiles for local regions containing multiple cells, this task requires modeling cell-to-cell expression variability, which is strongly structured by cell type. We propose Genomics-Guided Cell-Type-Specific Mixture-of-Experts (GC-MoE), which estimates cell-type probabilities with a routing network and softly combines cell-type-specific experts for gene expression prediction. To further encode cell-type-dependent gene programs, we introduce the Cell-Type-Specific Co-Expression-Aware Predictor (CAP), together with a lightweight Cell-to-Cell Interaction Attention (C2CA) module for neighboring-cell context. Experiments and ablations on public single-cell ST datasets show consistent improvements over existing single-cell and adapted spot-level baselines.

Large Language Models (LLMs) have revolutionized the field of Natural Language Processing (NLP) by automating traditional labor-intensive tasks and consequently accelerating the development of computer-aided applications. As researchers continue to advance this field with the introduction of novel language models and more efficient training/finetuning methodologies, the idea of prompt engineering and subsequent optimization strategies with LLMs has emerged as a particularly impactful trend to yield a substantial performance boost across diverse NLP tasks. To best of our knowledge numerous review articles have explored prompt engineering, however, a critical gap exists in comprehensive analyses of prompt optimization strategies. To bridge this gap this paper provides unique and comprehensive insights into the potential of diverse prompt optimization strategies. It analyzes their underlying working paradigms and based on these principles, categorizes them into 11 distinct classes. Moreover, the paper provides details about various NLP tasks where these prompt optimization strategies have been employed, along with details of different LLMs and benchmark datasets used for evaluation. This comprehensive compilation lays a robust foundation for future comparative studies and enables rigorous assessment of prompt optimization and LLM-based predictive pipelines under consistent experimental settings: a critical need in the current landscape. Ultimately, this research will centralize diverse strategic knowledge to facilitate the adaptation of existing prompt optimization strategies for development of innovative predictors across unexplored tasks.

The goal of coreset selection methods is to identify representative subsets of datasets for efficient model training. Yet, existing methods often ignore the possibility of annotation errors and require fixed pruning ratios, making them impractical in real-world settings. We present HyperCore, a robust and adaptive coreset selection framework designed explicitly for noisy environments. HyperCore leverages lightweight hypersphere models learned per class, embedding in-class samples close to a hypersphere center while naturally segregating out-of-class samples based on their distance. By using Youden's J statistic, HyperCore can adaptively select pruning thresholds, enabling automatic, noise-aware data pruning without hyperparameter tuning. Our experiments reveal that HyperCore consistently surpasses state-of-the-art coreset selection methods, especially under noisy and low-data regimes. HyperCore effectively discards mislabeled and ambiguous points, yielding compact yet highly informative subsets suitable for scalable and noise-free learning.

Early forecasting of clinical trials duration supports planning, optimized resource allocation, and cost control. Developing effective AI predictors remains challenging, requiring collaboration between medical and computational expertise. Although a publicly available dataset exists for this task, it is not well-suited for development of an AI-driven predictor capable of accurately predicting clinical trials duration. This limitation arises from its lack of critical features that significantly influence trials length, as well as the presence of incompletely and inappropriately selected trials. To empower development of AI predictors, we introduce the Clinical Trial Duration Prediction (CTDP) dataset, derived from 81,943 completed drug trials across 1,520 ICD-10 categories. The CTDP dataset contains 51 features on clinical trial characteristics including phase, purpose, condition, eligibility, enrollment, arm information, design, endpoint, site, sponsor, and summary that together influence trial duration. CTDP dataset development through collaboration between clinical trial experts and AI researchers will enable the creation of robust and accurate AI-driven applications for predicting clinical trial durations.

Ozone is an important secondary pollutant, affects numerous processes in the Earth's atmosphere and significantly contributes to air pollution, climate, and environmental change, while its presence at higher altitudes is essential for protecting life on Earth. Tropospheric ozone is a crucial component in the formation of photochemical smog, and its changes are linked to human activity. Recent advancements in tropospheric ozone monitoring help to perform quantitative analysis on global and regional spatiotemporal appearances, and also enable investigation of ozone formation mechanisms involved in the troposphere. Based on the literature, there remains substantial scope for the development of regional and global datasets as well as new predictors for ozone pollution tasks. Moreover, such analyses are needed across individual regions, which highlights an

TCAV (Testing with Concept Activation Vectors) is an interpretability method that assesses the alignment between the internal representations of a trained neural network and human-understandable, high-level concepts. Though effective, TCAV suffers from significant computational overhead, inter-layer disagreement of TCAV scores, and statistical instability. This work takes a step toward addressing these challenges by introducing E-TCAV, a framework for efficient approximation of TCAV scores, which is based on extensive investigation into three key aspects of the TCAV methodology: 1) the effect of latent classifiers on the stability of TCAV scores, 2) the inter-layer agreement of TCAV scores, and 3) the use of the penultimate layer as a fast proxy for earlier layers for TCAV computation. To ensure a solid foundation for E-TCAV, we conduct extensive evaluations across four different architectures and five datasets, encompassing problems from both computer vision and natural language domains. Our results show that the layers in the final block of the neural network strongly agree with the penultimate layer in terms of the TCAV scores, and the commonly observed variance of the TCAV scores can be attributed to the choice of the latent classifier. Leveraging this inter-layer agreement and the degeneracy of directional sensitivities at the penultimate layer, E-TCAV guarantees linearly scaling speed-ups with respect to the network's size and the number of evaluation samples, marking a step towards efficient model debugging and real-time concept-guided training.

Dataset distillation aims to compress large datasets into compact yet highly informative subsets that preserve the training behavior of the original data. While this concept has gained traction in classification, its potential for image Super-Resolution (SR) remains largely untapped. In this work, we conduct the first systematic study of dataset distillation for SR, evaluating both pixel- and latent-space formulations. We show that a distilled dataset, occupying only 8.88

Automated detection and classification of structural cracks and surface defects is a critical challenge in civil engineering, infrastructure maintenance, and heritage preservation. Recent advances in Computer Vision (CV) and Deep Learning (DL) have significantly improved automatic crack detection. However, these methods rely heavily on large, diverse, and carefully curated datasets that include various crack types across different surface materials. Many existing public crack datasets lack geographic diversity, surface types, scale, and labeling consistency, making it challenging for trained algorithms to generalize effectively in real world conditions. We provide a novel dataset, StructDamage, a curated collection of approximately 78,093 images spanning nine surface types: walls, tile, stone, road, pavement, deck, concrete, and brick. The dataset was constructed by systematically aggregating, harmonizing, and reannotating images from 32 publicly available datasets covering concrete structures, asphalt pavements, masonry walls, bridges, and historic buildings. All images are organized in a folder level classification hierarchy suitable for training Convolutional Neural Networks (CNNs) and Vision Transformers. To highlight the practical value of the dataset, we present baseline classification results using fifteen DL architectures from six model families, with twelve achieving macro F1-scores over 0.96. The best performing model DenseNet201 achieves 98.62

Road traffic accidents claim millions of lives annually, majorly involving vulnerable road users such as pedestrians and cyclists. Proactive traffic scene risk assessment with priority to vulnerable traffic actors can drastically reduce the chances of collision and save lives. To improve the safety of advanced driver-assistance systems (ADAS), we propose Object Risk Estimation (ORE), a novel framework that combines motion, lane-position, and class labels to enable context-aware and ethically prioritized risk assessment. It uses a multitask framework that integrates object detection, Time-to-Contact (TTC) estimation, and semantic segmentation in a single deep neural network. Further, it uses post-processing to combine these outputs to estimate risk per object and overall scene risk. Leveraging a shared backbone, ORE achieves a 27% reduction in inference time while delivering superior performance, with 1.5 higher mAP % and 11% lower oMiD loss on the Waymo dataset. Its modular design ensures transparency, with intermediate outputs directly explaining risk scores, eliminating the need for post-hoc analysis. By facilitating timely transfer of control and protecting vulnerable road users, ORE offers a practical, interpretable, and efficient solution for ADAS and autonomous driving safety, scalable to diverse traffic scenarios and higher autonomy levels.

"Learning by Teaching (LbT)" helps learners deepen their understanding by explaining concepts to others, with questions playing a vital role in identifying knowledge gaps and reinforcing comprehension. However, existing systems for generating such questions often rely on rigid templates and are expensive to build. To overcome these limitations, we developed a system using Large Language Models (LLMs) to create dynamic, contextually relevant questions for LbT. In our English vocabulary learning study, we examined which learner characteristics best leverage the system's benefits. Our results showed improved memory retention over traditional methods at three and seven days of testing, with ten participants. Additionally, we identified traits linked to better learning outcomes, highlighting the potential for tailored approaches. These findings support the development of scalable, cost-effective solutions to enhance LbT methods across various fields.

Early diagnosis of plant diseases is critical for global food safety, yet most AI solutions lack the generalization required for real-world agricultural diversity. These models are typically constrained to specific species, failing to perform accurately across the broad spectrum of cultivated plants. To address this gap, we first introduce the FloraSyntropy Archive, a large-scale dataset of 178,922 images across 35 plant species, annotated with 97 distinct disease classes. We establish a benchmark by evaluating numerous existing models on this archive, revealing a significant performance gap. We then propose FloraSyntropy-Net, a novel federated learning framework (FL) that integrates a Memetic Algorithm (MAO) for optimal base model selection (DenseNet201), a novel Deep Block for enhanced feature representation, and a client-cloning strategy for scalable, privacy-preserving training. FloraSyntropy-Net achieves a state-of-the-art accuracy of 96.38

Integrating cell segmentation with tracking is critical for achieving a detailed and dynamic understanding of cellular behavior. This integration facilitates the study and quantification of cell morphology, movement, and interactions, offering valuable insights into developmental processes, drug response, and disease mechanisms. Traditional segmentation and tracking methods rely on labor-intensive annotations—such as full segmentation masks or bounding boxes for every cell in each frame—which severely limit throughput and scalability. PointTrack addresses these challenges by introducing a weakly-supervised pipeline that leverages sparse point annotations in the initial frame to automatically generate high-fidelity cell masks and maintain consistent object identities across subsequent frames. PointTrack begins with multiple sparse point annotations per cell in the first image of a time-lapse sequence. An interactive segmentation engine uses these annotations to delineate accurate cell boundaries, eliminating the need for exhaustive mask drawing. A dedicated point-tracking module then propagates the annotated points through the entire video, adapting to changes in cell appearance and motion with minimal user intervention. By reducing annotation effort to a few clicks per cell while preserving mask quality and identity continuity, PointTrack significantly accelerates the preparation of large-scale microscopy datasets. Extensive evaluation on two diverse time-lapse datasets—CTMC and DeepCell—demonstrates robust segmentation and reliable tracking performance under varying imaging modalities, cell densities, and noise conditions. Qualitative examples highlight precise delineation of overlapping cells and seamless trajectory recovery through transient occlusions. With streamlined annotation requirements and automated processing, PointTrack offers an effective and scalable solution for large-scale microscopy-based cell segmentation and tracking.

Deep neural networks are among the most successful algorithms in terms of performance and scalability across different domains. However, since these networks are black boxes, their usability is severely restricted due to a lack of interpretability. Existing interpretability methods do not address the analysis of time-series-based networks specifically enough. This paper shows that an analysis in the frequency domain can not only highlight relevant areas in the input signal better than existing methods but is also more robust to fluctuations in the signal. In this paper, FreqAtt is presented - a framework that enables post-hoc interpretation of time-series analysis. To achieve this, the relevant frequencies are evaluated, and the signal is either filtered or the relevant input data is marked. FreqAtt is evaluated using a wide range of statistical metrics to provide a broad overview of its performance. The results show that using frequency-based attribution, especially in combination with traditional attribution on top of the frequency-optimized signal, provides strong performance across different metrics.

Multi-sensor ML models for EO aim to enhance prediction accuracy by integrating data from various sources. However, the presence of missing data poses a significant challenge, particularly in non-persistent sensors that can be affected by external factors. Existing literature has explored strategies like temporal dropout and sensor-invariant models to address the generalization to missing data issues. Inspired by these works, we study two novel methods tailored for multi-sensor scenarios, namely Input Sensor Dropout (ISensD) and Ensemble Sensor Invariant (ESensI). Through experimentation on three multi-sensor temporal EO datasets, we demonstrate that these methods effectively increase the robustness of model predictions to missing sensors. Particularly, we focus on how the predictive performance of models drops when sensors are missing at different levels. We observe that ensemble multi-sensor models are the most robust to the lack of sensors. In addition, the sensor dropout component in ISensD shows promising robustness results.

Automated chest X-ray report generation requires precise cross-modal grounding to ensure clinically reliable descriptions. However, existing vision-language models rely on implicit attention mechanisms that fail to enforce explicit region-word correspondence and disease-level consistency. We propose Game-Theoretic Alignment Network (GTA-Net), a vision-language framework that formulates report generation as a cooperative game-theoretic alignment problem. The model introduces a BinaryGameAligner that models interactions between image regions and text tokens using similarity-based payoff matrices with Shapley-inspired importance weighting. To enforce clinical semantics, we further develop a Disease-Aware Ternary Aligner, which captures joint interactions among images, reports, and structured disease concepts. GTA-Net combines a Swin-based visual encoder with a LoRA-adapted large language model and is trained with a unified objective for generation and alignment. Experiments on CheXpertPlus and IU-XRay demonstrate state-of-the-art performance across standard generation metrics and improved clinical consistency, highlighting the effectiveness of explicit game-theoretic alignment for medical vision-language generation.

Accurate segmentation of brain tumour sub-regions from multi-parametric MRI is critical for treatment planning yet remains challenging due to morphological variability, class imbalance, and overlapping appearances of tumour regions across imaging sequences. We propose SegGuidedNet, a three-dimensional residual encoder–decoder network introducing a novel SegAttentionGate module that explicitly supervises the decoder to produce spatially discriminative attention maps for each tumour sub-region necrotic core, peritumoral oedema, and enhancing tumour via a lightweight auxiliary loss, adding less than 0.2

BACKGROUND:Therapeutic decisions in clinical oncology are commonly established through interdisciplinary consensus in multidisciplinary cancer conferences (MCC). Artificial intelligence (AI) may support these processes by generating data-driven treatment recommendations (TR). We developed and evaluated an explainable AI system designed to reproduce MCC-based treatment decisions for metastatic and non-metastatic prostate cancer (PC). METHODS:Clinical data from patients with histologically confirmed PC discussed in MCC between 2015 and 2022 were transformed into structured datasets. A hierarchical modeling framework was implemented to first predict overarching treatment categories and subsequently specify therapeutic strategies. Multiple machine learning and deep learning algorithms were trained to replicate MCC recommendations. Model performance was assessed using F1-scores. RESULTS:A total of 5478 MCC cases including 76 clinical input variables and 23 treatment output parameters were analyzed. The AI system generated automated TR with high predictive accuracy across both hierarchical levels. For high-level categories, F1-scores reached 0.89 for surgery and 0.81 for radiation therapy. For detailed recommendations, F1-scores reached 0.99 for prostatectomy and 0.98 for PSMA-ligand therapy. Lower performance in anti-cancer drug categories likely reflects smaller sample sizes. Feature importance analyses ensured model transparency and interpretability. CONCLUSION:To our knowledge, this study presents one of the first large-scale explainable AI system capable of generating MCC-aligned treatment recommendations for metastatic and non-metastatic PC within a multi-target framework. It incorporates the largest reported number of clinical input and treatment output parameters in this setting. Strong predictive performance and interpretability support its potential as a scalable decision-support tool in multidisciplinary oncology. Prospective validation is warranted.

Effective document intelligence models rely on large amounts of annotated training data. However,procuring sufficient and high-quality data poses significant challenges due to the labor-intensive andcostly nature of data acquisition. Additionally, leveraging language models to annotate real documents raises concerns about data privacy. In this context, synthetic document generation has emergedas a promising, privacy-preserving alternative. We propose DocDjinn, a novel framework for controllable synthetic document generation using Vision-Language Models (VLMs) that produces annotated documents from unlabeled seed samples. Our approach generates visually plausible and semantically consistent synthetic documents that follow the distribution of an existing source dataset through clustering-based seed selection with parameterized sampling. By enriching documents with realistic diffusion-based handwriting and contextual visual elements via semantic-visual decoupling, we generate diverse, high-quality annotated synthetic documents. We evaluate DocDjinn as a source of trainingdata across eleven benchmarks spanning key information extraction, question answering, documentclassification, and document layout analysis. Our experiments show that with only 100 real training samples, our framework achieves on average 87% of the performance of the full real-world dataset. To our knowledge, this is the first work demonstrating that VLMs can generate faithful annotated document datasets at scale from unlabeled seeds that can effectively enrich or approximate real, manually annotated data for diverse document understanding tasks. We publicly release our code and 140k+synthetic document samples.