ABSTRACT Addressing the pain points of difficulty in obtaining expert strategies and insufficient generalisation of imitation learning in dynamic industrial edge environments, this paper proposes computational task offloading framework driven by multimodal generative adversarial imitation learning (GAIL-TO). GAIL-TO learns optimal strategy features from suboptimal historical logs through the adversarial mechanism of generator-discriminator, designs a spatiotemporal feature fusion encoder; constructs a lightweight adversarial training architecture, adapting to resource-constrained devices; develops an edge collaborative training mechanism, utilising server computing power to perform adversarial training, with only lightweight generators deployed at the terminal. It demonstrates superior performance compared to standard IL and MADDPG.

ABSTRACT Advancing AI enhances CPS capabilities but complicates human-system interactions. To address these challenges, this paper proposes a scenario-driven CPS design method based on an industrial-intelligence framework, focusing on system elements, hierarchies, and activities. We apply this framework to diverse scenarios—including power-system maintenance, flight management, and intelligent mining—using model-based techniques. These applications validate the framework’s effectiveness and provide a robust modeling foundation for real-world CPS development.

ABSTRACT The traditional unidimensional view of enterprise technology knowledge networks is challenged by distinguishing friendship networks, avoidance networks, and clique membership, each shaping knowledge quality and employee productivity differently. Rather than assuming knowledge flows are equal, this study shows avoidance networks foster misinformation and low-quality knowledge, while cliques create isolation and redundancy. Employees embedded in avoidance heavy or clique dominated structures encounter fragmented or outdated information, hindering enterprise system integration. Although friendship networks provide higher quality knowledge, they cannot fully offset avoidance and clique driven fragmentation. These findings reconceptualize enterprise system adaptation by highlighting network composition as a central determinant.

ABSTRACT Trust management (TM) underpins information security by addressing threats through governance, culture, and risk mitigation. This study applies the zero trust (ZT) paradigm, treating trust as a continuously verified structural asset. Examining cultural alignment, financial readiness, compliance, and operational change, it explores how enterprises embed trust in security strategies. Despite rising interest, enterprise-level ZT dimensions remain underexplored. Using principal–agent theory, qualitative interviews, entropy-based MCDM, and PLS-SEM survey analysis, the study finds cultural and financial factors most critical, supported by compliance and operations. It offers a conceptual framework and phased roadmap to strengthen TM, resilience, and regulatory alignment.

ABSTRACT Cryptocurrencies are increasingly the subject of fake news, increasing risks for market stability and investor decisions. To address this issue, we propose a multimodal framework to detect fake cryptocurrency news using text, image, and sentiment features with BERT, Swin Transformer, and RoBERTa, respectively. We use multi-head attention to combine these features to ensure the complementarity of features from different modalities. The fused representations are passed into a fully connected layer for final classification. Experimental results show that this framework achieves better accuracy and reliability than unimodal and multimodal models for detecting cryptocurrency misinformation.

ABSTRACT Digital transformation (DT) is reshaping construction, yet firms struggle to implement it. Drawing on the resource‑based view and upper echelons theory, we employ machine learning techniques to analyse an panel dataset comprising 861 annual observations from Chinese listed construction firms over a 12 year period from 2011 to 2022. We identify and rank DT drivers: a digital‑innovation mindset among top managers dominates, followed by inter‑organizational collaboration and boundary spanning; technology inputs and environmental support are secondary. We prioritize leadership and partnerships as levers for effective DT and offer practical guidance for governance and operations, aligned with UN Sustainable Development Goals.

ABSTRACT Supply chain financing can alleviate the financial constraints of small and medium-sized enterprises, blockchain can optimise supply chain financing strategies. The unique natural language processing capabilities of large language models have broad application prospects in supply chain financing decision-making. We innovatively introduce large language models to explore their application in blockchain-driven supply chain financing platforms. It is found that the excellent information processing capabilities of large language models significantly improves users’ financing efficiency. This study highlights the application potential of large language models and provides decision-making reference for managers.

ABSTRACT Human factors (HFs) in any digital healthcare systems (DHS) refer to the limitations and errors of human operators or users in the healthcare delivery and patient engagement. This review focuses on two situations (1) users are part of the system, (2) users are the client of the system. This paper aims to identify knowledge gaps in HFs in DHS and propose future research directions to close these gaps. The review discovered five key dimensions of HFs in DHS and nine relevant topic areas. From these, three major knowledge gaps and six future research directions are proposed to further advance the subject area of HFs in DHSs.

ABSTRACT This paper proposes a weighted mean filtering-based visual-inertial fusion SLAM system to address the growing demands of complex applications like inspection and rescue. The weighted mean filtering algorithm preprocesses the IMU data to improve the positioning accuracy and robustness. The visual-inertial SLAM system then serves as the experimental platform for angular error evaluation. Experimental validation on the Euroc dataset demonstrates our algorithm’s superior performance on tracking and mapping accuracy compared to other SLAM systems through multiple experiments and metrics. With CPU acceleration, the proposed system’s processing time is less than 10 ms, demonstrating its excellent real-time performance and its potential for practical applications.

ABSTRACT With rising demand for early English education, institutions face challenges in teacher training, high material costs, and limited tools suited for young learners. This paper proposes a human-centred generative AI system that leverages GANs to create visually engaging and pedagogically relevant content for children’s English picture books. By incorporating post-generation image feature extraction, the system ensures consistency and quality. It reduces production costs and enhances mobile learning, improving early language acquisition while boosting the competitiveness and credibility of English education providers.