[Purpose/Significance] Against the backdrop of rapid technological change and industrial transformation, China's labor market is undergoing profound restructuring, giving rise to multi-dimensional mismatches in the scale, structure, skills, and spatial distribution of talent supply and demand. Traditional governance models, which are largely dependent on statistical reporting and periodic planning, are increasingly inadequate in terms of timeliness, granularity, and evidence support. In response, this study focuses on matching talent supply and demand and integrates situation awareness theory with the intelligence-processing paradigm of "from information to decision." We propose a five-layer dynamic perception framework - "Data-Perception-Comprehension-Projection-Decision" - tailored to talent governance. The framework addresses the limitations of existing research, which often focuses on static measurements and neglects external events and policy shocks. It establishes a governance-oriented theoretical tool that meets major national strategic needs. By incorporating an event-driven dynamic sensing mechanism, the framework enables real-time identification and forward-looking regulation of talent mismatches in complex and evolving environments. [Method/Process] Theoretically, the study builds on Endsley's situation awareness model and further incorporates intelligence processes, causal inference, and complex systems theory, expanding the traditional "perception-comprehension-projection" cognitive sequence into a social closed loop governance system anchored by a robust data foundation and decision-feedback mechanism. Technically, we construct an event ontology and slot schema specific to talent supply-demand scenarios. Through event extraction and large language model-based semantic reasoning, the system generates structured "event-entity-relation" knowledge units from heterogeneous sources such as policy documents, industry reports, and job postings. Knowledge is then aligned and integrated using retrieval-augmented generation, domain knowledge graphs, and authoritative classification standards. At the modeling level, the framework integrates efficiency assessment, causal inference, and risk analysis into a unified "situation comprehension-situation projection" analytical architecture. Methods such as stochastic frontier analysis, difference-in-differences, and multi-agent simulation are employed to examine talent allocation efficiency, policy impacts, and supply-demand evolution. In application, a prototype system - combining data governance, event sensing, early warning, and decision support - is developed and tested using higher education major-industry matching as an empirical case. [Results/Conclusions] This research has developed an operational and verifiable theoretical and technical framework for the dynamic sensing of talent supply-demand alignment. It establishes a coherent chain that spans from "multi-source data integration" to "event-driven sensing," "causal mechanism analysis," "medium- and long-term trend projection," and "policy scenario evaluation," thereby facilitating a transformation in talent governance from static planning to dynamic monitoring, and from experience-based judgment to evidence-driven decision-making. The proposed system not only converts heterogeneous multi-source data into computable event knowledge, but also identifies the root causes of mismatches through causal inference and efficiency assessment. Moreover, it delineates potential future trajectories via trend forecasting and risk alerts. Building upon this foundation, decision-makers can leverage scenario simulation and strategy recommendations to obtain quantitative evidence and actionable plans for program restructuring, enrollment planning optimization, and curriculum reform. Nevertheless, the study has several limitations, including uneven data quality across sources, model dependence on historical samples, and the need for more fine-grained representation of complex behavioral mechanisms. Future research could build on the current work in several directions. For instance, it could incorporate richer data on individual behaviors and organizational strategies to enhance the model's ability to depict micro-level decision-making processes. It could also deepen the integration of causal inference with knowledge-graph techniques to improve the accuracy and clarity of identifying the effects of complex policy portfolios.

[Purpose/Significance] In the era of widespread social media, network cluster behavior has emerged as a significant phenomenon that shapes online public opinion and collective action. Although existing research has thoroughly examined macro-level drivers and developed evolutionary stage models for network cluster behavior, there is still a significant gap in our understanding of the micro-level cognitive mechanisms that dynamically propel its evolution. Cognitive biases, which are inherent tendencies in human cognition, are amplified in online group interactions. This study specifically addresses this gap by adopting a cognitive bias perspective to investigate the evolution mechanism of network cluster behavior. It is crucial to focus on campus hot events as highly relevant and sensitive case studies. These events often involve students, parents, educational institutions, and the wider public, covering core issues such as campus safety, management disputes, teacher-student relations, and student rights. Their inherent emotional resonance, rapid dissemination within specific online communities, and potential for severe damage to reputation and social order necessitate deeper understanding. The core innovation and significance of this research lie in: 1) Systematically integrating cognitive bias theory to analyze the complete lifecycle evolution of network cluster behavior in campus events; 2) Empirically revealing how specific biases dynamically manifest and interact at various stages, shaping the trajectory of network cluster behavior; 3) Providing a richer theoretical framework for network cluster action theory; 4) Offering empirical evidence for formulating targeted governance strategies to mitigate risks associated with campus-related online crises, thereby promoting constructive online discourse and campus stability. [Method/Process] To rigorously investigate the core research question, this study employed the grounded theory methodology. Based on sustained high popularity rankings on the "Zhiwei Shijian" platform, ten representative campus hot events were systematically selected to ensure coverage of diverse campus issues. Extensive datasets of user comments related to these ten events were collected from the Sina Weibo platform, serving as the core empirical foundation. The data collection timeframe spanned the complete lifecycle of each event, from initial emergence to eventual subsidence. Following the grounded theory process, the collected textual data underwent a meticulous three-stage coding procedure to induce and refine textual themes. Through this process, facilitated by qualitative data analysis software, a substantive theoretical model was ultimately constructed. This model delineates the evolutionary path and internal mechanisms of network cluster behavior in campus events under the influence of cognitive biases. The grounded theory method was deemed highly appropriate due to its capacity for deeply exploring complex social processes and emergent phenomena directly from rich, context-specific data. [Results/Conclusions] The study found that the evolution mechanism of network cluster behavior in the context of campus hot topics mainly consists of five stages: public opinion induction, public opinion bias, public opinion diffusion, public opinion outbreak, and public opinion subsidence. Based on these findings, governance strategies for such campus network events have been proposed, including identifying triggering factors, avoiding cognitive biases, enhancing user literacy, promoting collaborative guidance, and mitigating secondary risks.

[Purpose/Significance] In the digital age, data elements have become a key factor in production, while insufficient bargaining power in the supply chain poses significant operational risks to enterprises. How to leverage the opportunities of the digital economy, maximize the role of data elements, and avoid operational risks caused by insufficient discourse power in the supply chain has become a key issue that enterprises urgently need to address. Investigating how data utilization enhances this power is vital for building resilient supply chains and informing governance decisions. This method is also effective for further utilizing data elements. It provides micro evidence that helps us understand how data elements can optimize resource allocation and empower organizational decision making. [Method/Process] This study employs a rigorous, empirical approach using panel data from China's A-share listed companies from 2003 to 2022. A two-way fixed effects model serves as the primary estimator to control for unobserved heterogeneity. To credibly address potential endogeneity issues, such as reverse causality and sample selection bias, we implemented a comprehensive identification strategy. This methodology incorporates the use of instrumental variables, Heckman's two-stage correction model, and a series of robustness checks including alternative variable constructions and sub-sample analyses. Furthermore, we conducted mechanism analysis to elucidate the transmission channels and heterogeneity analysis to examine conditional effects across different types of firms. [Results/Conclusions] The empirical results demonstrate that the improvement of data element utilization level can effectively strengthen a firm's supply chain bargaining power and reduce the dependence of enterprises on large suppliers and customers, enhance their bargaining power and influence in the supply chain. This conclusion still holds true after robustness tests such as replacing the regression model, adding control variables, and adjusting the sample period. Mechanism analysis results indicate that the utilization level of data elements primarily empowers supply chain discourse through two channels: improving supply chain efficiency and alleviating financing constraints. Firstly, data elements optimize the inventory management, logistics scheduling, and supply chain collaboration of enterprises, improving operational efficiency and reducing dependence on key suppliers and customers. Secondly, data elements improve the information transparency of enterprises, reduce external financing costs, enhance the liquidity of funds, and make them more autonomous and bargaining power in supply chain transactions. A heterogeneity analysis revealed significant differences in the empowering effects of data elements among different types of enterprises. Among them, data elements have a more significant effect on enhancing the discourse power of supply chain for non-labor-intensive and non-asset-intensive enterprises, as well as a stronger promotional effect on non-technology-intensive and non-high-tech industry enterprises. This suggests that companies that rely less on traditional physical resources are better able to use data to gain a competitive advantage. This study establishes a robust theoretical basis for data-driven supply chain management and presents significant policy implications. One limitation is its focus on listed companies. Future research could expand this inquiry to include small and medium-sized enterprises and global supply chain contexts.

[Purpose/Significance] Open scientific data policies play a pivotal role in promoting the open sharing, unrestricted access to, and reuse of scientific data, thereby enhancing research efficiency and driving innovation. Despite their significance, research on the diffusion of these policies has predominantly focused on policy formulation, often neglecting the critical aspect of policy adoption and implementation at the local government level. This study aims to addres this gap by comprehensively examining the factors that influence the adoption of open scientific data policies by prefecture-level governments in China. The research was motivated by the need to understand how these policies spread across different regions, as well as the underlying mechanisms that facilitate or hinder their adoption. In doing so, the study expands the existing knowledge base by shedding light on the dynamics of policy diffusion in the context of open scientific data, a relatively under-explored area compared to other policy domains. [Method/Process] To achieve its objectives, the study employed an integrated research methodology. First, it utilized a policy diffusion model, adapted from the well-established Berry model, to theoretically frame the research. This model was enhanced by incorporating insights from a comprehensive literature review, which helps identify key internal and external factors influencing policy diffusion. Second, the study employed the event-history analysis to empirically test these factors using data from 286 Chinese cities over the period from 2018 to 2022. This method allows for the examination of the temporal sequence of policy adoption and the identification of causal relationships between the influencing factors and policy diffusion. Finally, a fuzzy-set qualitative comparative analysis (fsQCA) was applied to refine the understanding of multiple causal configurations that lead to successful policy adoption. This approach captures the complexity and interdependence of factors in policy diffusion processes, offering a nuanced perspective that goes beyond traditional statistical methods. [Results/Conclusions] The study identified four primary pathways for the diffusion of open scientific data policies in China: resource-driven, organization-and-human-capital-led, multi-stakeholder collaborative, and technology-guided. The resource-driven pathway emphasizes the significance of research funding and the establishment of professional organizations in facilitating policy adoption. The organization-and-human-capital-led pathway highlights the role of government official mobility and a skilled workforce in driving policy diffusion. The multi-stakeholder collaborative pathway underscores the importance of coordinated efforts among various stakeholders, including government agencies, research institutions, and industry partners. Last, the technology-guided pathway focuses on innovation capacity and professional management as key drivers of policy adoption. The findings reveal a heavy reliance on administrative measures in driving policy diffusion, which may lead to unintended consequences such as policy sustainability issues and a lack of alignment with local needs. Therefore, local governments are encouraged to adopt tailored diffusion strategies that consider their specific contexts and resource endowments. Future research should explore the performance of these policies in achieving their intended outcomes and conduct comparative studies across different regions to enhance the generalizability of the findings.

[Purpose/Significance] In the contemporary era of rapid technological advancement, understanding the intrinsic linkages between scientific research and technological innovation is critical for guiding strategic decision-making, optimizing resource allocation, and promoting effective technology transfer. Scientific publications and patents represent two complementary yet heterogeneous knowledge sources, with distinct linguistic styles, terminologies, and documentation structures, which often create a significant semantic gap. Traditional methods of linking scientific and technological (S&T) knowledge rely primarily on lexical overlap, keyword co-occurrence, or citation analysis. These methods are limited in their ability to capture deeper semantic relationships, particularly across non-homologous texts. To address this challenge, this study proposes a novel approach leveraging large language models (LLMs) for data augmentation, aiming to uncover latent semantic associations between research paper topics and patent topics. The key innovation of this work lies in using LLMs not merely for text generation but as a semantic bridge to enhance cross-domain knowledge alignment, thereby advancing the methodological toolkit for science-technology linkage studies. This approach offers potential contributions to knowledge mapping, thematic analysis, and strategic innovation management, particularly in areas where domain-specific terminology or conceptual divergence hampers conventional analyses. [Method/Process] The proposed method employs ChatGPT-4 as a knowledge-enriched intermediary to generate semantically enhanced textual variants of existing S&T documents in the energy-saving domain. Specifically, the LLM was used to perform synonym-based paraphrasing, expansion, and semantic inference on research paper abstracts and patent summaries, producing augmented texts that retain domain relevance while highlighting latent semantic connections. These enhanced texts were used to extract features that were subsequently incorporated into a non-patent citation prediction task, which serves as a practical evaluation of the method's effectiveness. By comparing predicted associations against existing citation links, the study assesses the capacity of LLM-derived features to capture cross-domain topic relatedness beyond lexical similarity. The approach relies on the theoretical premise that LLMs can model high-level semantic patterns, enabling the inference of conceptual correspondence even when explicit terminology differs between scientific and technological texts. [Results/Conclusions] The experimental validation process involved four baseline models, and it was found that features derived from the augmented texts consistently improved prediction performance. The area under the ROC curve (AUC) increased by 13.91%, 16.90%, 16.21%, and 15.69% across the four models, respectively, demonstrating the efficacy of LLM-based data augmentation in bridging the semantic gap between S&T knowledge. These results suggest that the method can uncover latent topic associations, facilitate cross-domain term alignment, and support knowledge discovery tasks that conventional lexical-based approaches may overlook. However, the study is limited by its focus on a single application domain, leaving open questions regarding generalizability across multiple S&T fields. Future work should extend the methodology to diverse domains, investigate the robustness of the LLM-generated semantic bridges, and explore automated mechanisms for scaling cross-domain knowledge integration. Overall, this research provides a promising framework for enhancing the semantic connectivity of heterogeneous knowledge sources. This contributes to a broader understanding of the interactions between science and technology and informs data-driven strategies for managing research and innovation.

[Purpose/Significance] In today's knowledge economy, where scientific research and innovation drive social change, accurately and scientifically assessing the social impact of scientific research achievements has become key to optimizing the global scientific research ecosystem. This article focuses on the social impact evaluation system of the international scientific research achievement. It provides in-depth analysis of typical international models and strategic guidance for China to build a more comprehensive and efficient evaluation system. [Method/Process] Based on the theoretical definition of the social impact of scientific research achievements, eight major cases of third-party evaluations were selected: the EU SIAMP, the US STAR METRICS, the UK REF, the Dutch SEP, the Italian VQR, the Canadian CAHS, the Australian ERA, and the Japanese NIAD-QE. Using a cross-national comparative analysis method, a comprehensive analysis was conducted across three dimensions: system elements (establishment time, establishing entity, main characteristics, evaluation scope, and strategic objectives), mechanism processes (definition of evaluation objects, establishment of evaluation procedures, application of evaluation results), and methodological tools (definition of social impact-related content, evaluation methods, and indicator content). Subsequently, relevant information was collected through literature research and online research to identify key characteristics. [Results/Conclusions] International evaluation systems are guided by national strategic needs and incorporate social impact into the entire research lifecycle management process through legislation. These systems also link influence to funding allocation. These systems operate using policy-driven mechanisms, collaborative efforts among stakeholders, data-driven methodologies, and dynamic feedback loops. The key characteristics of typical international research evaluation models are as follows: 1) Multi-dimensional indicators: Moving beyond traditional academic metrics, evaluation frameworks now encompass a wide range of impacts, including the effects of research outcomes on social welfare, industrial development, and employment. 2) Dynamic adjustment: As the socio-economic and technological environment evolves, the social impact evaluation systems of international research outcomes also undergo dynamic adjustments and innovations. 3) Multi-stakeholder collaboration: This involves diversified participation, cross-disciplinary and cross-departmental collaboration, and the full involvement of stakeholders throughout the process. Based on the above findings, this study offers insights at different stages of social impact assessment of scientific research achievements. Prior to implementation, additional indicators aligned with domestic strategic priorities, such as environmental sustainability, social equity, and cultural heritage preservation, should be incorporated alongside traditional metrics, and the policy and legal framework should be refined. During implementation, a multi-stakeholder collaborative evaluation platform should be established, and a dynamic system incorporating resilience coefficients should be developed to address uncertainties. After completion, a long-term monitoring and tracking mechanism should be implemented to understand ongoing impacts, with feedback-driven updates to the indicator system. This approach aims to foster a healthy evaluation ecosystem, accelerate the translation of research outcomes into societal value, and promote the integrated development of scientific research and social progress.

[Purpose/Significance] With the rapid advancement of artificial intelligence (AI), university libraries are undergoing a deep transformation from traditional resource repositories to intelligent service ecosystems. This transformation poses a significant challenge to the conventional competencies of librarians and underscores the necessity for a systematic reconstruction of these competencies. Existing studies often lack empirically supported and integrative models, and they seldom bridge the gap between AI application and competence development. To address these shortcomings, this study proposes a core competence model for hybrid AI librarians, integrating technical, service, and management dimensions. The research highlights its innovation by not only theorizing but also empirically validating the model through grounded data, positioning the study as a meaningful contribution to the discourse on digital librarianship. Different from previous literature, it integrates AI platform practices within the competency framework. This integration serves to enrich both theoretical underpinnings and enhance the practical applicability of the theory. This provides actionable implications for the sustainable development of librarianship in the context of national strategies for digital transformation and technological innovation. [Method/Process] The study employed a mixed-methods approach. First, a literature review was conducted to analyze trends in AI applications within university libraries. Then, semi-structured in-depth interviews were carried out with ten librarians from multiple universities that have deployed the DeepSeek intelligent platform. The participants covered technical, service, and management positions, with more than three years of experience using AI tools and a distribution across middle to senior professional titles. Following data collection, the grounded theory was applied with three levels of coding (open, axial, and selective) to inductively derive categories and explore how technical, service, and management competencies interact. The principle of data saturation was strictly observed to ensure methodological rigor, and no additional categories emerged after the three competency domains were established. [Results/Conclusions] Findings indicate that the core competencies of hybrid AI librarians revolve around three interdependent domains. Technical competence involves intelligent tool operation, data analysis, and system maintenance, supporting the integration of AI into daily workflows. Service competence emphasizes user-centered design, personalized recommendation, and human-AI collaborative interaction, ensuring that technical functions translate into user value. Management competence addresses resource allocation, cross-department collaboration, and ethical governance, safeguarding sustainability, compliance, and innovation. Together, these dimensions form a "technology-service-management" dynamic balance model, characterized by reinforcing loops in which technology drives service, service demands managerial support, and management stabilizes technology-service integration. Furthermore, a training and cultivation framework was proposed, offering differentiated professional pathways based on librarians' roles and growth stages. The study concluded that such a model not only enhances service effectiveness but also contributes to national innovation strategies. The study's limitations include its scope, which is limited to a single country and a small sample size. Future research should expand the sample base, employ comparative studies across institutions, and further examine the weighting of competencies.