3020 papers
This paper proposes a multi-tenant edge-cloud hybrid LLM serving architecture that combines W4A16-quantized speculative decoding with asynchronous communication and a multi-tenant verification orchestrator to simultaneously address privacy, latency, and resource utilization challenges.
This paper introduces JPAL-HA, a safe learning model for safety-critical environments that leverages human action justifications to generate hypothetical queries, thereby significantly improving safety and sample efficiency while reducing human training burden.
This study demonstrates that using explainable machine learning to analyze item-level safety climate data reveals nonlinear, threshold-dependent behavioural risk patterns driven by management responsiveness, training, and experience, challenging traditional linear aggregation models and offering a more precise basis for proactive accident prevention.
This paper proposes a sparse ℓ0-regularized quadratic surface support vector machine (QSVM) to address overfitting and interpretability issues in kernel-free nonlinear classification, introducing a penalty decomposition algorithm with provable optimality and convergence guarantees that demonstrates competitive performance and sparsity on both benchmark and real-world credit datasets.
This paper presents a PRISMA-based systematic review revealing the limitations of current auditory-centric solutions for deaf learners and proposes an AI-driven multimodal communication framework integrating speech recognition, natural language processing, and sign language generation to enable real-time visual support and achieve meaningful inclusion in classrooms.
This paper reveals that standard scaffold-based evaluation methods introduce a hidden allocation bias by incorrectly treating chemically heterogeneous "no-scaffold" molecules as a single unit, which distorts generalization metrics across common benchmarks, and proposes an "Empty-Scaffold-Aware" repair to restore balanced representation and accurate performance assessment.
This paper introduces C3Q-MARL, a connectivity-aware framework that integrates graph-theoretic topology control with a QAOA-inspired optimization module into multi-agent reinforcement learning to enhance UAV swarm coordination, collision avoidance, and connectivity preservation under partial observability and communication constraints.
This study presents the first scale-by-scale benchmark of YOLOv11 and YOLOv12 variants on the BCCD dataset, revealing that while their overall mean performance is nearly identical, YOLOv12 excels at the nano scale for latency-critical applications whereas YOLOv11 outperforms it at medium and large scales, indicating that model selection should be driven by specific scale and recall requirements rather than a uniform assumption of superiority.
This paper introduces RLHOARNet, a MONAI-based 3D U-Net architecture featuring a lightweight residual boundary refinement module that achieves state-of-the-art performance and real-time inference speeds for multi-modal brain tumor segmentation on the BraTS 2020 dataset.
FedHyperM is a privacy-preserving federated learning framework for edge devices that addresses modality heterogeneity by employing hypergraph-based prediction and modality-sensitive aggregation to significantly outperform existing baselines in accuracy and F1-score.