2731 papers
This paper proposes a multi-task learning framework featuring a novel residual neural dimension reductor with temporal and channel attention (RNDR-TCA) to jointly estimate multiple cardiovascular parameters from blood pressure waveforms, effectively mitigating negative transfer and outperforming existing single-task and multi-task models in accuracy and robustness.
This paper proposes SPA, a topology-independent, hop-by-hop routing protection algorithm that ensures seamless, incremental deployment and guarantees protection against all single-failure scenarios with minimal path stretch, outperforming existing solutions like ESCAP, U-turn, and NPC.
This paper proposes a multimodal diagnostic framework utilizing Audio Spectrogram Transformers and vision-language models to analyze respiratory sounds and patient metadata, demonstrating superior accuracy and interpretability compared to traditional physician auscultation and previous CNN-based approaches for asthma diagnosis and remote monitoring.
The H-XAI-Cervix study demonstrates that a conditional hierarchical EfficientNet-B3 model significantly outperforms flat classifiers in distinguishing cervical cell classes on the Herlev dataset, though this advantage was not observed on the SIPaKMeD dataset where performance approached a near-ceiling level.
This paper presents a comprehensive benchmarking study on the NIH ChestX-Ray14 dataset demonstrating that a Swin Transformer V2-B framework, enhanced by specific strategies for class imbalance and threshold optimization, outperforms existing state-of-the-art deep learning models in multi-label chest disease classification.
This paper benchmarks four multi-agent reinforcement learning architectures against traditional dispatching rules in a stochastic OSAT environment, demonstrating that while learned policies significantly improve throughput, flow time, energy efficiency, and OEE compared to FIFO, they do not uniformly outperform the SPT rule, thereby establishing a reproducible framework for evaluating short-horizon dispatching performance.
This paper introduces the LLM-Serving Scheduler Portability Benchmark (LSSP) to demonstrate that while scheduling policy rankings show strong agreement across some workload sources, they exhibit significant variability and limited portability across different operating regions and evaluation metrics, necessitating that scheduler comparisons be interpreted as conditional on their specific experimental context.
This study evaluates BitkiKormacı, a Turkish small language model system for plant protection, demonstrating that while retrieval-augmented generation significantly improves performance on specific development benchmarks, the system still exhibits critical failures in safety and resistance tasks, underscoring the necessity for rigorous independent expert review before practical deployment.
This paper introduces the Agentic Enterprise Architecture Framework (AEAF), a formal extension to traditional enterprise architecture metamodels that incorporates a dedicated agency layer with stratified Datalog semantics to systematically model, analyze, and govern non-human AI agents by defining their authority, accountability, and coordination constraints within the enterprise structure.
This paper proposes GAT-MAPPO-EIG, a Graph Attention Multi-Agent Proximal Policy Optimization framework that leverages deep reinforcement learning to efficiently solve large-scale, dynamic escape interdiction games by learning coordinated interception strategies without relying on computationally expensive traditional optimization methods.