4157 papers
This paper presents µDCN, a high-performance, ML-orchestrated Named Data Networking (NDN) architecture for edge environments that integrates eBPF/XDP, Rust-based QUIC, and TensorFlow Lite to dynamically adapt packet granularity, thereby improving throughput and latency while reducing resource consumption compared to existing implementations.
This paper critically evaluates an adaptive inference-time reasoning method using sparse guide queries, finding that while it significantly reduces full-token injections and shows promise in distinguishing signal, it ultimately fails to deliver statistically significant improvements in generation accuracy or calibration over a frozen base model.
This paper proposes a novel machine learning method for solving interface problems that combines neural network-based subspace approximation with geometry-aware quadrature and an a posteriori error estimator to achieve high accuracy and automatically handle singularities.
This paper introduces a benchmark and evaluation protocol for merged cyber-attack incident repositories that reveals how supervised imputation of missing attributes creates significant, often concealed, performance inflation, while demonstrating that entity embeddings and a source-agnostic three-part protocol are essential for obtaining reliable multi-attribute attack forecasts.
This paper presents a resource-efficient, DSP-free, and fully pipelined hybrid PRNG for the Xilinx Zynq-7000 platform that combines Xoshiro128**, Weyl sequences, and an integer Tent map to achieve high throughput (5.82 Gbps) and pass rigorous statistical tests while maintaining a small footprint suitable for embedded applications.
This paper introduces State-inferer, a white-box framework that integrates program slicing, large language models, and formal model checking to accurately extract and verify protocol state machines from source code, achieving superior transition-level accuracy and uncovering four previously unknown security vulnerabilities in TLS, IKEv2, and SSH implementations.
This study challenges the misconception that declining publication shares indicate field decline by demonstrating that while the overall AI in education literature grew fivefold from 2016 to 2026, specific subfields experienced differential growth rates driven by "substrate portability"—the adaptability of research constructs to new data environments—rather than functional obsolescence.
This paper introduces LoRA-Perturb, an efficient perturbation-based framework that quantifies layer-wise importance to enable differentiated rank and learning rate allocation in LoRA, significantly reducing computational overhead while achieving state-of-the-art performance on various large language model benchmarks.
This paper evaluates FusionNet Lite, a compact 1,420-parameter federated learning model for predictive maintenance, demonstrating its superior efficiency and conditional performance under label heterogeneity and temporal generalization while highlighting significant variability in metrics and the need for further industrial validation.
This paper introduces EMM, an education-centered benchmark designed to evaluate generative AI's ability to simultaneously recognize manipulative advisor utterances and generate proportionate first-person student responses, revealing a critical gap between accurate manipulation detection and appropriate autonomous replies.