2833 papers
The paper presents HEAL, a bilingual English-Luganda RAG chatbot that outperforms commercial models in delivering accurate, guideline-compliant disease surveillance information to frontline health workers in Uganda by combining GPT-4 with a fine-tuned translation model.
This paper proposes SubNM, a technology-node-aware hardware Trojan detection framework that leverages multi-library netlist synthesis and modular graph neural networks to achieve near-perfect detection accuracy and robust generalization across diverse semiconductor process nodes, overcoming the limitations of single-configuration evaluation methods.
The paper introduces Agora, a framework that differentiates multi-agent LLM systems through six continuous psychological drives mapped to model parameters, demonstrating that this approach significantly enhances performance on analytical tasks compared to baselines like AutoGenStyle, while simpler coordination strategies remain more effective for creative and short-format generation.
This paper presents a generalized, correlation-aware probabilistic search framework that leverages semantic segmentation as a soft detector for multi-target robotic search, introducing a Bayesian belief-update mechanism based on the Generalized Binomial Distribution to handle altitude-dependent observation correlations and validated through simulation and real-world UAV landing-site detection.
This paper proposes a hardware-efficient Android malware detection framework for resource-constrained IoT devices that transforms executables into 1D opcode signals to extract hybrid features, achieving high classification accuracy with ultra-compact Decision Tree and Neural Network models quantized for deployment on ARM-Cortex-M4 microcontrollers with minimal memory footprint and nanosecond-level inference latency.
This paper proposes a Behavior-Driven Reinforcement Learning (BRL) framework that enhances secure routing in multi-hop IoT networks by integrating local forwarding behavior into autonomous decision-making, thereby achieving superior packet delivery, lower overhead, and improved energy efficiency compared to existing protocols without relying on centralized control or cryptographic overhead.
This paper presents a voice-interactive embodied AI robotic scrub technician that successfully demonstrates autonomous surgical instrument handoffs and take-backs with 81% overall success, highlighting its potential for operating room integration while identifying grasp acquisition as a key area for future improvement.
This paper proposes the CICS-TCMN framework, which integrates causal structure learning, a Contextual Intervention Consistency Selector, and a Temporal Causal Memory Network to deliver accurate, temporally aware, and clinically interpretable personalized cardiovascular risk predictions.
This research demonstrates that fine-tuned large language models, particularly LLaMA-3.2-1B and the parameter-efficient DistilBERT, can achieve near-perfect accuracy in detecting XSS, SQL, and command injection attacks, offering a highly effective solution for mitigating these critical web security threats.
This paper presents a deep learning-based hate speech detection system for the low-resource Afan Oromo language using XLM-RoBERTa and Afro-XLMR models, demonstrating that the fine-tuned XLM-RoBERTa achieves superior performance with 95.3% F1 score and 95.8% accuracy on a dataset of 22,000 social media samples to enhance online safety and social cohesion in Ethiopia.