2608 papers
This paper demonstrates that open-weight small and medium-sized local LLMs can effectively and efficiently automate Q-matrix generation for Cognitive Diagnosis Models, offering a privacy-preserving and resource-accessible alternative to costly expert curation and closed-source frontier models.
This paper presents an AI-driven, end-to-end data analytics framework integrating Python ETL, MySQL modeling, and interactive dashboards to analyze retail transactions, revealing that while gender spending parity exists despite volume skew, current promotional discounts and subscription models fail to significantly drive basket size or per-visit spend, thereby highlighting critical gaps in acquisition and monetization strategies.
This paper proposes a comprehensive, multi-layered framework that integrates BERT-BiLSTM-Attention encoding, semantic embeddings, n-gram fingerprinting, and stylometric analysis to robustly detect exact matches, paraphrased plagiarism, mosaic copying, and AI-generated content while identifying inconsistent authorship within documents.
This paper introduces SOC Copilot, a lightweight, CPU-only, unsupervised multi-model anomaly detection engine that fuses an improved Isolation Forest and a Deep Autoencoder with SHAP-based explainability to achieve 99.84% accuracy in analyzing HDFS security logs without requiring labeled data or GPU infrastructure.
This paper introduces Cretogram, a new dataset of 11,611 manually annotated isolated epigraphic characters from Archaic and Classical Crete, and evaluates its utility for computational recognition using HOG-SVM and deep learning models, achieving a top macro-F1 score of 95.53% with ResNet-18 while highlighting specific character confusions inherent to the script's local variations.
This study demonstrates that in breast-cancer histopathology segmentation, the measured contributions of small architectural components in ablation studies can vary significantly in magnitude and even change direction across repeated training runs, highlighting the critical need for replication before drawing definitive conclusions about component efficacy.
This study demonstrates that uncontrolled stochastic variations in nominally identical deep learning runs for brain-tumour MRI classification can exceed the performance differences between distinct model architectures, thereby undermining the reliability of standard ablation studies and necessitating stricter reproducibility protocols.
This systematic review synthesizes research on explainable machine learning methods utilizing multimodal educational data to enable the early detection and transparent explanation of mathematics learning difficulties, while highlighting current challenges and proposing a future research agenda focused on privacy, standardization, and ethical deployment.
This paper introduces AETHEL, an open-source and reproducible framework that systematically audits the trustworthiness of clinical machine learning models by evaluating their discrimination, calibration, explainability, robustness, and clinical utility across heterogeneous healthcare settings to address the challenges of domain shift.
This paper introduces the Asymptotically-Embedded Physics-Informed Neural Network (AE-PINN), a mesh-free deep learning framework that overcomes the spectral bias of standard PINNs by embedding asymptotic expansions of boundary layers into the network architecture, thereby achieving high accuracy and stability in solving unsteady singularly perturbed convection-diffusion-reaction equations with sharp gradients.