2661 papers
This paper presents the design, implementation, and initial validation of a scalable online platform that leverages generative AI to deliver personalized digital skills training to youth, demonstrating positive user feedback and functional performance while recommending further longitudinal research to assess broader educational impact.
This study demonstrates that a compact desktop workstation can effectively host open-weight language models for journal editorial tasks, revealing that while performance does not strictly correlate with model size, combining these models with simple deterministic checks achieves near-perfect detection of guideline violations.
This paper proposes "Cardio Fusion," a dual-stream regularized deep learning architecture that integrates optimized tabular data processing with imaging/ECG analysis to achieve robust, generalizable cardiovascular disease prediction across diverse and heterogeneous clinical datasets.
This paper proposes and evaluates a layered cybersecurity assurance architecture for autonomous deep-space multi-agent systems that utilizes machine-verifiable evidence, provenance, and dynamic reputation management to ensure the integrity and trustworthiness of AI-generated scientific claims in the absence of real-time human oversight.
This paper presents UAS-OS, a novel operating-system-level control plane that governs autonomous AI workloads through Agent Control Blocks and zero-copy memory management, achieving robust intent control and fault isolation with sub-50 microsecond overhead suitable for real-time mission-critical environments.
This paper proposes a multi-dimensional collaborative enhancement framework for YOLOv5 that integrates LAB-space CLAHE, a high-resolution P2 detection layer with Varifocal Loss, and deformable convolution with channel attention to effectively address low-light, small-target, and geometric deformation challenges in industrial visual inspection, achieving superior performance over mainstream detectors on a specialized ton bag dataset.
This paper proposes DBDNet, a dual-branch decoupling network that leverages wavelet transform and a frequency-based physical prior to effectively separate modality-shared low-frequency backgrounds from modality-specific high-frequency details, thereby achieving state-of-the-art performance in multi-modality image fusion and downstream object detection.
DeepCork is a two-stage deep learning framework that combines a ResNet152-based binary classifier with a YOLO-based detector to achieve high-precision, real-time defect inspection of natural cork stoppers while effectively balancing localization accuracy, latency constraints, and predictive uncertainty in industrial manufacturing.
BioPretext is a novel multimodal self-supervised learning framework that unifies geometric imaging and sequential biological signals through cross-modal alignment and dynamic fusion to enhance orthopedic feature extraction and downstream task performance.
This paper presents HoneyMed, a protocol-aware honeypot framework for IoMT networks that simulates medical services to capture and classify attacks in real time using machine learning, achieving 98.50% accuracy with Random Forest while generating blockchain-ready evidence without risking production assets.