2608 papers
This paper introduces GridWorld3DEnv, a reproducible 3D voxel-based simulation benchmark with standardized metrics and evaluation protocols, to address the lack of fair cross-study comparisons in multi-UAV cooperative coverage by systematically evaluating MARL algorithms like QMIX and VDN while releasing all code and datasets.
This paper proposes a bidirectional Path Integral Monte Carlo algorithm enhanced by Multiple Importance Sampling to efficiently estimate quantum circuit transition amplitudes in extremely sparse path spaces, demonstrating superior convergence and scalability for circuits with up to 4096 qubits compared to unidirectional approaches.
This paper evaluates local credit assignment methods for CPU-based Transformers, finding that while asynchronous readout-gradient consensus improves training throughput, both it and predictive coding approaches fail to match backpropagation's quality or establish a general, quality-preserving acceleration alternative.
This paper introduces AIRouteOpt, a trust-aware deep policy-routing framework that leverages Proximal Policy Optimization to dynamically select next hops based on link quality and recursively updated trust scores, significantly improving packet delivery rates in tactical vehicle MANETs under joint mobility and routing-plane attacks compared to traditional trust-blind and threshold-based approaches.
This paper introduces TurkCuisineBench, a 72-item, source-grounded short-answer benchmark that evaluates large language models' factual knowledge of Turkish cuisine and heritage, demonstrating that semantic evaluation significantly improves accuracy assessment over exact-string matching while highlighting substantial performance variations across different model endpoints.
This study demonstrates that nearly half of clinical trial eligibility criteria for non-small-cell lung cancer involve logical disjunctions rather than simple conjunctions, revealing that the common assumption of independent requirements leads to significant safety and efficiency errors in automated patient-trial matching.
This study introduces the UzHistQA dataset and demonstrates that while retrieval-augmented generation significantly reduces hallucinations in Uzbek history question answering, enforcing citation and abstention mechanisms is necessary to completely eliminate fabricated answers, highlighting the need to separately evaluate retrieval quality and generation faithfulness in under-resourced languages.
This study demonstrates that lightweight convolutional neural networks, particularly MobileNetV2 and EfficientNetB0, offer a computationally efficient and reliable solution for automated brachial plexus localization in ultrasound images, making them well-suited for resource-constrained point-of-care environments despite achieving performance comparable to the heavier ResNet34 architecture.
This paper presents and validates the Lifecycle Cyber-Resilience Architecture (LCRA) through large-scale simulation, demonstrating that integrating supply-chain assurance, identity-centred access governance, and explainable adaptive operation significantly enhances the containment of malicious incidents and operational stability in high-risk IoT environments compared to baseline approaches.
This paper proposes a transparent "Glass-Box" forensic framework utilizing forty interpretable stylometric metrics to demonstrate that, despite lexical convergence, distinct and widening divergences in information-theoretic predictability, structural properties, and natural variability persist between human and AI-generated text across diverse datasets, offering a robust foundation for authorship verification.