2786 papers
This paper presents a critical survey of 183 studies (2018–2026) on graph learning for cold-start and data-scarce recommendation, offering a taxonomy-driven analysis of methodologies and highlighting the urgent need for stricter evaluation protocols and more reliable side-information integration to address the current unevenness in evidence.
The paper proposes SymCLIP, a novel framework that enhances multi-intent recognition in social media images by introducing Symmetric Vision–Language Prompting to synergistically couple visual and textual features and Dynamic Label Integration to enable adaptive semantic representation, achieving state-of-the-art performance on the Intentonomy dataset.
This study presents a locally deployable, knowledge graph-based intelligent guidance system for anesthesiology residency training that utilizes a dual-engine retrieval-augmented generation framework to deliver accurate, guideline-compliant, and privacy-preserving educational support without relying on cloud-based large language models.
This study demonstrates that for crop recommendation tasks, a well-tuned single Random Forest model achieves accuracy comparable to complex ensemble methods, suggesting that simpler, cost-effective solutions are sufficient without the need for redundant model combinations.
This paper presents a Design Science Research study on a closed-domain Smart HR system for SMEs that demonstrates a Corrective RAG adaptation with same-chapter hard-negative training significantly improves answer faithfulness over standard RAG, despite a trade-off in overall accuracy.
This paper introduces SCOPE, an uncertainty-aware, training-free framework that reconstructs coherent attack behaviors under incomplete observations by integrating behavior-centric grouping, distributional TTP mapping, and Top-K Viterbi sequence inference to outperform existing baselines in accuracy and robustness.
This paper presents a highly optimized, fine-tuned EfficientNet-B0 model using transfer learning that achieves 97.57% accuracy in automatically detecting nine types of maize diseases from real-world images, offering a computationally efficient tool to assist smallholder farmers in early disease identification.
This paper presents a resource-efficient visual navigation system for UAVs that utilizes panoramic neural compression and dual-snapshot fusion to enable continuous, closed-loop route following and homing with minimal memory overhead and high positional accuracy.
This paper proposes an optimized AI-based intrusion detection system that utilizes GridSearchCV to tune six machine learning and deep learning models on the CICDDoS2019 dataset, demonstrating that Random Forest and XGBoost achieve superior accuracy and efficiency in detecting various DDoS attacks compared to other algorithms.
This paper proposes Q-LDPSO, a novel quantum-enhanced particle swarm optimizer featuring Quantum Population Initialization and a Quantum Bidirectional Search Strategy, which achieves superior feature selection and intrusion detection performance (99.63% accuracy) for IoT environments when combined with an RF-XGBoost ensemble classifier.