2833 papers
This study reveals that widely used language model agents exhibit a robust in-group trust bias based on arbitrary labels—a subtle social dynamic that remains undetectable by standard aggregate behavioral audits because it manifests in *who* receives actions rather than *which* actions are chosen.
This paper introduces CHRONOS, a novel hybrid architecture that formalizes scientifically defensible counterfactual history as a constrained structural-causal-model problem over temporal knowledge graphs, while transparently reporting honest limitations in scalability and quantum optimization performance observed during preliminary synthetic testing.
This study demonstrates that enterprise firewall decisions can be near-deterministically reconstructed from traffic logs using supervised learning, revealing that high model accuracy often reflects the replication of specific policy proxies and context-dependent rules rather than independent threat detection, thereby enabling retrospective policy audits while highlighting significant limitations in cross-environment generalization and uncertainty calibration.
This study demonstrates that a hybrid AI framework combining Convolutional and Recurrent Neural Networks with LLM-based data augmentation and content generation can significantly optimize client acquisition on social media, achieving high accuracy in classifying target groups and promotional posts within VKontakte communities.
This paper presents a systematic literature review on the integration of Generative AI and Digital Twins for predictive maintenance within Industry 5.0, analyzing current trends and challenges while proposing a novel Data-Model-Human framework to guide future human-centric research in smart manufacturing.
This paper introduces SA2A, an intent-driven system agent architecture that enforces mutual zero-trust by validating all responses at the requester's entry gate, thereby achieving robust security against diverse attacks and automatic resilience to provider changes with only minimal latency and bandwidth overhead.
This study proposes an evidence-graded educational knowledge graph and a trustworthy recommendation framework for the Sichuan-Chongqing region that integrates fine-grained content extraction, risk gating, and learner state estimation to deliver verifiable, cross-subject K-12 resource recommendations while explicitly managing the trade-off between traceability and coverage.
This paper proposes an AI-native Manufacturing Operating System (M-OS) for autonomous smart factories that integrates Digital Twins, Multi-Agent Systems, Physics-Informed Machine Learning, and Reinforcement Learning into a unified six-layer architecture to synchronize decision-making across scheduling, maintenance, and quality control, while outlining a mathematical framework and experimental protocol for validation without presenting empirical results.
This meta-analysis demonstrates that misconfigurations in foundational telemetry services (AAA, NTP, and Syslog) within SME networks critically undermine incident response capabilities by causing forensic timeline collapse, loss of non-repudiation, and visibility blindness, thereby highlighting the necessity of using risk-free virtual sandboxes to proactively identify and mitigate these systemic risks.
This paper introduces ARQM, an Automated Requirements Quality Management tool that leverages modern AI and Machine Learning methods alongside Explainable AI to perform requirement screening and quality analysis based on the IEEE 29148:2018 standard, demonstrating the potential of these technologies to learn general patterns while visualizing detected violations.