2786 papers
This study demonstrates that integrating LoRA-adapted Stable Diffusion with ControlNet for synthetic data augmentation modestly enhances YOLOv8s-based underwater garbage detection performance, though its effectiveness varies by object class and depends on the quality of generated images and annotation alignment.
This paper presents a comprehensive pipeline for Arabic propaganda detection that demonstrates how GPT-4.1-driven few-shot data augmentation significantly outperforms traditional methods, enabling a fine-tuned BERT model to achieve state-of-the-art results (F1-Micro=0.66) that surpass both previous benchmarks and few-shot LLM approaches, while revealing that linguistic features are detrimental and data augmentation is the critical success factor in this low-resource, imbalanced setting.
CandidateFusion-MKAN is a unified framework that achieves robust single-output multimodal crisis classification by integrating modality-supported candidate-set learning, candidate-set likelihood, and bounded adaptive soft targets to effectively handle diverse supervision, missing or degraded evidence, and conflicting or complementary cues.
This paper introduces HoKuVit, a scalable Convolutional Vision Transformer that integrates Kuramoto-based convolutional and Hopfield-based linear layers into an Oscillatory Neural Network architecture, achieving state-of-the-art performance on CIFAR-10 while demonstrating a viable pathway for scaling ONNs to deep AI pipelines.
This paper introduces Geometry-aware LegONet (gLegONet), a modular operator learning framework that decouples physical mechanisms from specific geometries by pretraining variational blocks on an ambient spectral domain and adapting them to arbitrary target domains through boundary-defined affine manifolds, thereby enabling reusable PDE solving and law discovery without geometry-specific retraining.
This paper introduces the Saudi Legal Corpus for AI, a comprehensive, officially sourced, and auditable article-level dataset of 290 Saudi legislative instruments featuring unique provenance tracking, structured analytical layers like citation graphs and defined-term glossaries, and a retrieval benchmark demonstrating the superior performance of metadata-enhanced search over standard baselines for Arabic legal NLP.
This paper addresses the challenge of lexical ambiguity in Sanskrit-English machine translation by constructing a sense-augmented parallel corpus using LLMs and a custom sense inventory, demonstrating that explicitly incorporating word sense information significantly improves translation quality across different model architectures.
This paper presents the design and implementation of a fully automated smart restaurant system that integrates mobile applications, IoT infrastructure, and service robots to eliminate human intervention, resulting in significant improvements in wait times, order accuracy, and delivery speed while achieving excellent user usability.
This paper introduces UBET, a training-free one-shot segmentation framework that prioritizes constructing closed-region hypotheses before identity arbitration using frozen proposals and analytic DINOv2 evidence, achieving state-of-the-art performance across diverse benchmarks without requiring model training.
This study demonstrates that a deep learning model can accurately estimate heart rate at the 2 mmol·L⁻¹ lactate threshold across running, cycling, rowing, and kayaking, outperforming traditional methods within specific sports but requiring sport-specific validation due to reduced generalization accuracy across different disciplines.