10 September 2026, Volume 46 Issue 9
-
Select all|
-
LIBRARY TRIBUNE. 2026, 46(9): 1-14.With the rapid development of emerging technologies such as big data and artificial intelligence（AI）,data ethics issues have become prominent,giving rise to the new discipline of data ethics. Reviewing the evolution of knowledge structures in this field both domestically and internationally,this paper identifies its core research topics：data privacy protection,algorithmic fairness,and the ethical risks of AI decision-making. Through an in-depth analysis of the challenges posed by emerging technologies,the study outlines future directions of data ethics. Furthermore,it examines the disciplinary attributes and positioning of data ethics and explores the practical pathways for establishing an independent knowledge system of data ethics with Chinese characteristics in a global context.
-
LIBRARY TRIBUNE. 2026, 46(9): 15-25.This paper aims to explore the relationship between the early indicators of preprints and the academic impact of journals during the transition from preprint to formal publication,with a view to providing new perspectives on assessing the potential value of preprints,the development of preprint platforms under open science,and journal publishing models. Focusing on preprints from the bioRxiv platform that were subsequently published,this study employed Spearman’s correlation analysis to examine the correlation among 17 variables,selected the appropriate indicators as independent variables,and then performed a multiple linear regression analysis on the relationship between preprint early indicators and journal academic impact. Among the preprint early indicators,six were found to have a positive impact on journal academic impact：Twitter mentions,number of Mendeley readers,number of Dimensions citations,TRiP peer reviews,PDF downloads,and interval;and only one indicator—full-text views—showed a negative impact. In the context of open science,the study proposes strategies for improving preprint platforms and optimizing journal publishing models,including emphasizing the development and utilization of preprint platforms,attaching great importance to social media engagement,and integrating peer review and PDF download features on these platforms.
-
LIBRARY TRIBUNE. 2026, 46(9): 26-40.While GAI-assisted academic reading improves efficiency,it also gradually leads university students to become increasingly over-reliant on intelligent technologies,subtly eroding their capacity for independent thinking. As a result,a learning pattern of “using without thinking” is becoming the norm. This study aims to elucidate the formation mechanism of this new type of cognitive inertia,thereby providing both theoretical and practical guidance for the appropriate use of GAI in academic contexts. Using a grounded theory approach,this study integrates the UTAUT2 model with relevant theories from cognitive psychology to develop a theoretical model. The model is empirically tested through structural equation modeling and then refined through follow-up interviews. The findings show that the use of GAI,when reinforced by emotional attachment and usage habits,can easily lead to technological dependence,which in turn induces new cognitive inertia. Meanwhile,metacognitive ability acts as a “brake”,mitigating the negative effects of technological dependence. In terms of teaching,universities should establish a “cognitive regulation-oriented” model for GAI usage. As for GAI tool design,it is recommended to promote “cognitive scaffolding-oriented” learning support,so as to achieve a human-centered digital transformation in education in which intelligence serves as an aid,not a replacement.
-
LIBRARY TRIBUNE. 2026, 46(9): 41-57.In the context of advancing open science and evolving academic evaluation systems,this study examines how code availability influences the disruptiveness of academic papers. Focusing on artificial intelligence（AI）,it seeks to determine whether open-source code enhances the innovative output of scientific research and helps identify breakthrough studies,thereby expanding the quantitative metrics of scholarly assessment. The study measures the disruptiveness of academic papers using the Disruption Index（D-index）and develops a high-dimensional fixed-effects regression model for empirical analysis based on a sample of top AI conference papers classified by the China Computer Federation（CCF）. Model specifications include controls for traditional bibliometric indicators to ensure the robustness of the results. Empirical findings show that code availability has a significant and positive effect on paper disruptiveness,though the strength and direction of this relationship vary across conference types. Furthermore,corporate involvement and the number of authors exert moderating effects,while the degree of interdisciplinarity plays a partial mediating role. Overall,the large language model-enhanced bibliometric analysis indicates that code availability promotes the disruptiveness and innovation potential of scientific research;however,this effect is conditional on organizational and collaborative contexts,and is partially mediated by cross-disciplinary integration mechanisms.
-
LIBRARY TRIBUNE. 2026, 46(9): 58-69.Generative AI is profoundly reshaping scientific research paradigms,bringing unprecedented opportunities for academic research. This paper employs the procedural grounded theory approach to code and analyze data collected through semi-structured interviews. Using the Triadic Reciprocal Determinism as the core analytical framework,it introduces the construct of outcome expectancy and proposes a behavioral mechanism model that explains how generative AI empowers academic research activities. The findings indicate that the application of generative AI in academic research activities is jointly driven by the external environment（technological diffusion,research atmosphere）and individual cognition（intrinsic motivation）. Its application manifests across three dimensions—innovation,optimization and decision-making—and presents a dual-impact pattern characterized by the coexistence of "intelligent leaps" and "multidimensional challenges." This study reveals the underlying structure of academic research activities in the context of human-AI collaboration,elucidates the behavioral characteristics and mechanism through which generative AI empowers such research,thereby providing a reference for the development of intelligent research and the establishment of academic behavior norms from the perspective of human-AI integration.
-
LIBRARY TRIBUNE. 2026, 46(9): 70-79.Given the challenges that increasing interdisciplinary integration poses to knowledge services in university libraries,there is an urgent need to explore and redefine the role of librarians in this process. Taking a typical interdisciplinary project in marine scientific research as a case study,this paper examines the knowledge service demands of various user groups through semi-structured interviews and identifies five types of knowledge needs. It also summarizes four typical characteristics of librarian engagement in such projects：the immediacy of knowledge needs,the interdisciplinary nature of the research,the diversity of service content,and the complexity of user communities. Based on these findings,the study develops a framework for understanding the role of librarians in interdisciplinary research projects. It then proposes the following recommendations for role transition：moving beyond fixed-position perceptions,actively engaging in decision-making,functioning as a coordinator,returning to the all-round librarian role,and enhancing knowledge and skills.
-
LIBRARY TRIBUNE. 2026, 46(9): 80-93.To address the limitations of traditional entity-relation extraction methods that rely heavily on large-scale annotated data and model fine-tuning,resulting in high deployment costs,this study explores a parameter-free large language model（LLM）-based information extraction paradigm to enhance structural stability and extraction reliability in scientific literature. We propose a structure-enhanced extraction approach based on prompt engineering for LLMs. The method integrates structure-preserving syntactic splitting,dynamic few-shot example selection based on semantic similarity,structured output constraints,and an original-text backtracking matching verification mechanism. In addition,both pipeline and joint extraction strategies are incorporated to guide LLMs toward high-quality entity and relation extraction. Experimental results demonstrate that the proposed method significantly improves the information extraction performance of LLMs. Specifically,Grok-4.1-fast（Pipeline）achieves the best overall performance,with F1 scores of 0.8703,0.5786,0.5712,and 0.7764 on NER,Rel,Rel+,and RE in the SciER dataset,and 0.7939,0.4920,0.4855,and 0.7352 in the SciNLP dataset,respectively. These findings indicate that,without fine-tuning,LLMs can achieve stable and efficient entity-relation extraction from scientific literature,providing a feasible paradigm for low-cost information extraction.
-
LIBRARY TRIBUNE. 2026, 46(9): 94-104.In the context of evolving open science and scholarly communication systems,academic libraries face multiple challenges such as financial constraints,insufficient resource accessibility,and inadequate safeguards for research equity. Therefore,exploring open investment models,a crucial mechanism for driving the transformation of library collections and building a fair and sustainable knowledge ecosystem,is of great significance. By developing an integrated framework of "four-stage evolution and fourfold shift," the study delineates the evolutionary path of collection investment from Open Publication Funds（OPFs）and Transformative Agreements（TAs）toward Fair Open Access（Fair OA）and Open Infrastructures（OIs）,which in turn drives a systematic shift in collection development in terms of nature,architecture,content,and format. Drawing on international experiences,the study analyzes the practical constraints in China and proposes corresponding strategies tailored to local conditions. The findings suggest that libraries should adopt strategies such as developing a roadmap for reducing Article Processing Charges（APCs）,selecting investment channels based on a multi-dimensional evaluation system,and implementing a progressive budget reallocation to achieve the strategic transformation from resource purchasers to investors and co-builders within the open science ecosystem.
-
LIBRARY TRIBUNE. 2026, 46(9): 105-118.In the context of high-quality resources being channeled to the grassroots,an analysis of the current supply-demand alignment in public reading services for county residents and the proposal of targeted strategies can shed light on improving the effectiveness of public library services at the county level and supporting grassroots cultural development. Focusing on the public libraries in 20 key supported counties of Anhui Province,this study follows three principles（sufficient literature basis,grassroots-context adaptability and theoretical support）to identify the factors affecting the supply-demand matching of public reading services. Based on questionnaire data,it examines the current supply-demand status of these services through factor analysis,difference analysis,and visual analysis. The results indicate that the overall supply-demand matching of reading services at county-level public libraries is relatively satisfactory;however,the alignment of supply and demand for certain indicators needs to be further improved. Readers of different ages,educational backgrounds,and occupations have different service needs. Accordingly,targeted strategies are proposed from the cognitive,functional,and emotional perspectives.
-
LIBRARY TRIBUNE. 2026, 46(9): 119-135.Under the new paradigm of systematic conservation of cultural heritage,documentary heritage is no longer merely supplementary to other forms of cultural heritage. There is an urgent need to re-examine its protection and utilization,as well as its foundational value within the framework of systematic cultural heritage conservation. The development of national cultural parks serves as a key arena for implementing systematic conservation,providing a crucial institutional context for rethinking the functional boundaries and value structure of documentary heritage.Grounded in this paradigm,this paper examines the practice of building the Great Wall National Cultural Park. Drawing on multi-source textual materials collected through online research and fieldwork,it employs textual content analysis to define the specific conceptual connotations of the Great Wall-related documents as an independent category of cultural heritage. The study further identifies 13 roles and 39 functions that this heritage assumes across the Park’s four primary functional zones. On this basis,it establishes a value cognition system for Great Wall documentary heritage structured around "target zoning,multiple roles,and functional adaptation". It also reveals the internal logic underlying the transformation of Great Wall documentary heritage from an auxiliary resource to a foundational element for both the holistic development of national cultural parks and the systematic conservation of cultural heritage,thereby addressing the emerging practical needs of this paradigm.
-
LIBRARY TRIBUNE. 2026, 46(9): 136-146.To address the problems such as high subjectivity and low efficiency in manual damage grading,as well as the over-reliance of classical deep learning models on high-quality annotated data,this paper proposes a new approach for the automated damage assessment of ancient books using a multimodal large model. By converting the Standard for Damage Grading of Ancient Books and Special Collections（officially,Standard for Distinction of Disrepair of Ancient Books and Special Collection）into machine-understandable quantitative metrics,it presents a refined prompt-driven multimodal large model for visual semantic analysis,introduces a color reference chart to assist judgment,and ultimately generates structured JSON grading data. Experimental results show that on the Guji1000 dataset,which contains 1,000 samples,the method achieves an overall grading F1 score of 0.83. Ablation experiments confirm that the introduction of knowledge quantification significantly enhances the evaluation performance of the large model.
-
LIBRARY TRIBUNE. 2026, 46(9): 147-160.Against the backdrop of rapid advancements in generative artificial intelligence,large language models（LLMs）face challenges in specialized,knowledge-intensive fields such as ancient books,including knowledge lag,semantic deviation,and hallucinations. Focusing on the “Er”（the character for “Child”）fascicle of Yongle Dadian,this paper explores methods for developing ancient book knowledge bases driven by Agent Skills. First,LoRA fine-tuning and Retrieval-Augmented Generation（RAG）are employed to build a knowledge base for Yongle Dadian. Then,an agent architecture integrating a large language model,RAG,and domain-specific skills is designed. In the empirical study,a multimodal large language model is used to directly extract structured knowledge from ancient book images,and the combination of RAG and Agent Skills enables structured knowledge organization,semantic retrieval,intelligent question answering,and visual analysis. The results show that,compared to the LoRA fine-tuning and RAG methods,the Agent Skills-driven framework offers clear advantages in accuracy,hallucination control,and knowledge presentation. It enhances the computability and interpretability of ancient book knowledge organization,providing a new technical pathway for the knowledge-oriented development and intelligent services of ancient books.
