2731 papers
This paper introduces Qaamuuska-NLP, a structured 46,314-entry machine-readable Somali lexicon extracted via a deterministic rule-based pipeline from the 2012 monolingual dictionary *Qaamuuska Af-Soomaaliga*, which preserves detailed grammatical and lexical information while retaining original entry text for verification.
This paper introduces Mirage Mesh, a lightweight, multi-layer cyber deception framework for Linux environments that integrates network, system, software, and data-layer decoys with a unified event schema and automated pipeline to detect and analyze malicious activity resembling legitimate administration, while proposing a new metric to evaluate its effectiveness across the cyber kill chain.
This systematic review of ten high-risk-of-bias studies reveals that large multimodal models in clinical medicine are vulnerable to input-based corruption across four distinct failure classes, a threat that persists regardless of model characteristics or tested defenses and does not require an adversarial attacker to compromise output integrity.
This paper establishes a representation-independent tradeoff theorem demonstrating that erasing higher-order semantic information retained after lower-order fixation in unrestricted deterministic pushdown realizations necessitates a physical cost quantified by source stack exposure depth and canonicalization debt, with sharp lower bounds derived from the interplay between retained information and limited observation capacity.
This scoping review maps existing validation methods for Large Language Models in healthcare, categorizing them by task and evaluation type to provide stakeholders with a structured framework for selecting context-appropriate strategies before clinical implementation.
This systematic mapping study reviews 35 primary studies to analyze how artificial intelligence is transforming oral history from static archives into dynamic, interactive fields, highlighting significant advances in transcription and analysis while emphasizing critical ethical challenges and the necessity of trauma-informed, human-centered approaches.
This paper proposes a communication-efficient FedProx-Binary FedVote framework for dual-task Wi-Fi fingerprint indoor localization that significantly reduces uplink bandwidth by transmitting 1-bit update directions while maintaining high accuracy and robustness under strong Non-IID data conditions through FedProx regularization.
This paper demonstrates that while LLMs can be used to evaluate legal accuracy in low-resource languages like Quebec French, their reliability is insufficient for high-stakes deployment without human oversight, as even the strongest models exhibit moderate agreement, fail to consistently catch dangerous errors, and benefit significantly from step-by-step reasoning and ensemble methods.
This study proposes a physics-informed dual-view YOLO framework that integrates deep learning-based object localization with Magnus-effect aerodynamic modeling to significantly improve the accuracy of spinning tennis ball trajectory prediction and landing-point estimation under varying lighting conditions.
This study proposes an adaptive neural threshold network that outperforms logistic regression in prioritizing SEC filings for accounting enforcement review by learning context-dependent thresholds, while simultaneously establishing a framework of governance requirements to ensure the system's use as a confidential triage aid rather than legal evidence.