2662 papers
This paper demonstrates that optimizing reinforcement learning agents for emotion-related tasks naturally gives rise to internal reward signals that align with the core characteristics of biological emotions, supporting the principle that emotions evolve to alter actions in pursuit of intangible goals.
This study systematically benchmarks four traditional machine learning models against four transformer-based models across five public datasets, demonstrating that while transformer architectures like RoBERTa significantly outperform traditional methods in fake news detection accuracy, they incur substantially higher computational costs, thereby offering critical trade-off insights for model selection under varying resource constraints.
This paper formally defines the previously undocumented Nepali card game Dahal Jeet and presents a comprehensive computational benchmark demonstrating that search-based agents outperform learning methods, while revealing that trump length is a stronger predictor of victory than scoring cards and that controlling for deal variance is essential to accurately measuring human-agent performance differences.
This study demonstrates that for frozen pedestrian detectors under Gaussian noise and JPEG acquisition, selecting denoisers based on degradation-model matching is critical for detection recovery, whereas relying on traditional restoration fidelity metrics like PSNR or SSIM can lead to suboptimal choices.
AlamX is a privacy-preserving, fully local clinical intelligence platform that combines a Random Forest-based symptom and vital triage system with a locally executed biomedical LLM to provide accurate, uncertainty-aware disease predictions and conversational health guidance without transmitting sensitive user data to remote servers.
A within-subjects scenario study with U.S. adults found that users perceive time-synchronized, integrated data architectures for AI-based digital mental health systems more favorably than traditional siloed frameworks across system quality, information quality, and net benefits, though these modest perceptual advantages warrant further validation through functional prototype testing.
This paper proposes a novel asynchronous vertical federated learning framework integrated with blockchain technology to overcome the limitations of synchronous training and centralized coordination, thereby enhancing efficiency, security, and participant engagement through decentralized consensus, transparent model evaluation, and smart contract-based incentive mechanisms.
This paper introduces Bipolar Logic Networks (BLN), a rule-based machine learning model that achieves competitive accuracy with state-of-the-art interpretable methods while uniquely providing verifiable, interrogatable Boolean logic and calibrated probabilities essential for auditing high-stakes decisions.
This paper proposes a novel, stateless deterministic computing architecture that integrates Algorithmic Topological Resonance theory, a virtual 1-million-qubit processing unit, and reversible thermodynamic logic to overcome the memory wall, Shannon limits, and Landauer's thermodynamic barrier while ensuring hardware security and long-term digital permanence.
This study demonstrates that while machine learning models can achieve high dementia detection performance on the OASIS-2 dataset when rigorously evaluated with participant-level cross-validation to prevent data leakage, their apparent reliance on MRI features is inconsistent and often overshadowed by the Mini-Mental State Examination (MMSE), highlighting the critical need for independent validation to ensure clinical explainability and generalizability.