4159 papers
This paper proposes an intrinsically interpretable insurance fraud detection framework that combines prototype learning with concept bottleneck models to provide faithful, human-understandable explanations while achieving competitive performance on noisy claim data.
The paper introduces PRISM, a five-layer framework that integrates differential privacy, real-time temporal updates, cross-platform interoperability, and community-aware governance to optimize ethical, adaptive social network advertising, demonstrating through simulation that it significantly outperforms static single-platform baselines in user adoption while maintaining reach under privacy constraints.
The paper introduces HealthJev, an open-source, self-evolving agent that utilizes a lightweight, self-trained filtering mechanism (JevGate) to safely accumulate and distill longitudinal memory from medical encounters, significantly outperforming existing frameworks across diverse clinical tasks in a 90-day simulated benchmark.
This paper presents a comprehensive benchmark comparing three pipelines for PDE forecasting from sparse noisy observations, demonstrating that combining surrogate inference with numerical forecasting (A1) consistently reduces error compared to pure surrogate approaches while highlighting the significant impacts of training data volume, implementation choices, and numerical propagation on accuracy and computational cost.
This paper proposes SDGSA, an enhanced Gravity Search Algorithm incorporating a self-adaptive attenuation factor and a distance-fitness evaluation strategy to overcome premature convergence and improve performance in complex optimization and UAV path-planning tasks.
This paper introduces Bipolar Logic Networks (BLN), a neural-symbolic architecture that enhances causal discovery in high-stakes domains by generating fully auditable, condition-dependent mechanism candidates with explicit necessary and sufficient roles, achieving high structural recovery and stability across diverse datasets where traditional graph-based methods fall short.
This study demonstrates that five-fold ensemble disagreement serves as a validated, cost-free reliability metric for nnU-Net-based vertebral segmentation in spine CT, successfully flagging approximately 76% of poorly segmented vertebrae for review while requiring only a 9.4% review rate, with ensembling proving more effective than data expansion for improving cross-domain performance.
This paper presents a memory-safe, tiered eBPF and GPU-accelerated data plane for 5G networks that successfully detects adaptive BPFDoor backdoors by leveraging an immutable kernel socket invariant, outperforming traditional hand-written CUDA implementations while maintaining zero memory-safety violations under adversarial LLM attacks.
This exploratory study demonstrates that the "Synthetic Unconscious" patterns previously attributed to recursive self-interrogation in large language models are largely artifacts of a leading interview schedule, as a neutral control condition eliminated key findings like dual-component moral reasoning and priestly self-labeling while preserving only modest features like narrative self-description.
This paper proposes TinyML-CoDE, a comprehensive co-design and evaluation framework that addresses five critical research gaps in resource-constrained TinyML development by integrating standardized benchmarking, hardware-software co-design, lifecycle management, security, and toolchain interoperability to enable scalable and reproducible edge intelligence deployments.