2661 papers
This paper argues that to ensure a resilient and sustainable future for AI within a co-evolving noosphere, the field must shift from top-down representational alignment methods that enforce moral homogenization to an enactive alignment framework that preserves cognitive diversity through endogenous, participatory normativity.
This paper proposes the STG-MADAC framework, an attention-enhanced multi-agent reinforcement learning approach that utilizes spatiotemporal graph attention and multi-objective advantage decomposition to achieve superior decentralized weapon-target assignment in dynamic aerospace defense systems.
The paper presents MS-RTDETR, a real-time end-to-end detector that improves small-object detection by unifying multi-scale feature enhancement and query refinement through a novel Scale-Coupled Feature–Query Refinement (SCFQR) mechanism and an uncertainty-aware matching objective, achieving superior performance and robustness across diverse datasets without requiring auxiliary inference branches.
This study develops and empirically validates an "Auditable AI Ethics" competency framework that bridges high-level regulatory mandates with practical K-12 and university curricula, demonstrating through a pilot program that outcome-based, action-oriented pedagogy effectively transforms students from passive AI consumers into responsible creators capable of critical verification, bias mitigation, and legal compliance.
This paper introduces a coherent Quantum Feedforward Neural Network (QFNN) that utilizes trainable projector overlaps and alternating reflections to implement exact Chebyshev polynomial activations, thereby establishing rigorous theoretical foundations for its expressivity, universal approximation capabilities, and generalization bounds while demonstrating competitive performance on benchmark image classification tasks.
This paper demonstrates that poor open-world malware detection performance on the BODMAS benchmark is caused not by labeling errors or detector flaws, but by the presence of closely related families in the training set that render held-out "novel" samples actually familiar to unsupervised novelty scoring methods.
ParaWeb is a TypeScript library that implements ten parallel programming patterns across message passing, shared memory, and WebGPU compute shaders for Node.js and browsers, demonstrating significant speedups over sequential JavaScript and competitive performance against C++ libraries while highlighting the critical impact of data transfer overhead on GPU efficiency.
This paper proposes S3HNet, a stage-aware spectral-spatial hierarchical network that effectively segments ground-level urban hyperspectral images by preserving band-sensitive spectral evidence in shallow encoder stages and reconstructing spatially consistent semantic representations in the decoder, thereby outperforming existing dense segmentation models on benchmark datasets.
The paper introduces CRISP, a compiler-guided approach that enables safe polynomial-level parallelism for FHE-encrypted machine learning inference by trading runtime SIMD vectorization for compile-time OpenMP parallelism, resulting in significant speedups and latency reductions within the ANT-ACE compiler framework.
This paper proposes ADE-MDA-Net, an asymmetric dual-encoder network incorporating SE-dual pooling fusion, multi-dilation depthwise attention, and multi-scale spatial attention fusion modules to achieve accurate and computationally efficient pixel-level segmentation of challenging bearing surface defects.