最新刊期
卷 12 ， 期 2 ， 2026
-
- 摘要：As a pivotal strategy in the data era, the national integrated data market is intrinsically coupled with the process of marketization and value realization of data factors. By integrating the current practical landscape of China’s data market with the unique characteristics of data factors and market dynamics, it posits that the national integrated data market comprises four structural layers: rules, infrastructure, ecosystem, and governance. Furthermore, it proposes a strategic pathway that includes four breakthrough points: expanding supply and demand, strengthening institutional leadership, fostering ecosystem cultivation, advancing infrastructure construction, and improving the market governance system. It aims to facilitate the high-level allocation of data factors and the high-quality development of the digital economy.关键词：high-quality development;data factor334|87|0更新时间：2026-03-15
- 摘要：China's data factor market holds immense potential, yet the inefficiency in data circulation and utilization has hindered the full release of this potential, posing a core bottleneck that must be addressed in fostering a national integrated data market. The article, from the perspective of data circulation and utilization, is guided by the fundamental requirements of "five unifications and one opening" for the construction of a unified national market. It addresses the current challenges in the data market, such as inconsistent rules, non-interoperable facilities, uncoordinated entities, and an underdeveloped industry. The article proposes an overall approach that involves unifying institutional rules, coordinating infrastructure, fostering market consensus, and accelerating industry empowerment. Furthermore, it elaborates on specific implementation paths in four aspects: refining circulation and utilization standards, strengtheningcirculation and utilization infrastructure, improving data valuation and pricing mechanisms, and bolstering the data circulation and utilization industry.关键词：national integrated data market;cultivation path174|59|0更新时间：2026-03-15
- 摘要：Data has evolved into a fundamental strategic resource for developing new quality productive forces and driving high-quality economic growth. Cultivating a national integrated data market is a critical measure for overcoming the current inefficiencies caused by the fragmented allocation of data elements and for constructing new national competitive advantages. This paper elucidates the intrinsic nature and strategic necessity of the national integrated data market. It deeply dissects the multidimensional impediments currently facing this transition, including institutional barriers, infrastructure gaps, market ecosystem imbalances, and the absence of fiscal regulation. Based on the overarching framework of "Five Unifications and One Opening-up", the paper proposes a systemic reshaping path encompassing "institutional supply, infrastructure connectivity, ecosystem regulation, and governance synergy". Specifically, the paper advocates for: establishing a full-chain rights confirmation and standardization system; developing intensive and efficient data circulation infrastructure; fostering a layered, collaborative, and service-oriented market ecosystem; and innovating full-lifecycle fiscal and tax incentive mechanisms. The aim is to contribute a systemic institutional design scheme for building a unified, open, and orderly competitive data element market.关键词：data market;data institution;market construction;digital economy153|81|0更新时间：2026-03-15
- 摘要：Building a national integrated data factor market is a key measure to promote the development of the digital economy. The subject matter of data transactions underpins the development of the data market, and establishing a standardized, consistent, and unified subject matter for data transactions is the basic requirement for the national integrated data factor market. The differences between the data factor market and the data product market were sorted out, noting that most listings on existing data exchanges were data products rather than data factors. Data factors serve as the foundational raw materials for data factories, which must source them from the data factor market. Design specifications and standardization frameworks for data factor transaction subject matter were proposed, aiming to provide both theoretical support and practical guidance for accelerating the development of the national integrated data market.关键词：data market;subject matter of transaction;integration;digital economy192|67|0更新时间：2026-03-15
- 摘要：The efficient management and interoperability of digital resources serve as the cornerstone for realizing data empowerment. Traditional resource organization methods, centered on metadata management, are increasingly unable to adapt to the complex collaborative demands across different departments, hierarchies, and business domains, thereby limiting the deep utilization of digital resources. Data products are key elements of circulation in the data market, and unified identification serves as their foundation; research and practice in this area are critical for facilitating data circulation and sharing. Based on a systematic review of the research progress in identification coding methods, this paper addresses key challenges such as the growing complexity of data analysis demands, the heterogeneity of multi-source data, and semantic interoperability required for multi-task coordination in emerging data agent models. Itproposes a unified identification coding method for data products. By introducing machine-understandable high-dimensional semantic features, this approach enables data products to serve as a unified interface for agents to recognize and operate data, offering methodological reference and a standardization pathway for advancing the standardized, intelligent, and collaborative development of digital resource systems.关键词：unified identification;data agent;data circulation;digital resource129|61|0更新时间：2026-03-15
- 摘要：High-quality datasets determine the training performance of artificial intelligence models. The lack of a unified standard form and a quality-controllable, process-based production method for high-quality datasets has led to their insufficient supply and inefficient circulation, which has become a bottleneck restricting the development and application of artificial intelligence. From the perspective of data products, this paper proposes a five-tuple form for high-quality dataset products. Supported by full-link technical capabilities, we design a production process for such products and propose a product-oriented quality control method that covers the entire production chain. This work provides a theoretical basis and feasible solutions for the large-scale production and circulation of high-quality dataset products.关键词：data product;form;production process;data quality277|60|0更新时间：2026-03-15
TOPIC
-
Checkpoint accessing performance optimization method for the deep learning model training process AI导读
摘要：As LLMs become more widely used and their scale continues to expand, currently LLMs training faces issues such as high error rate and poor performance of checkpoint accessing. This paper reviews the strengths and weaknesses of existing methods for optimizing checkpoint accessing performance and introduces a novel method for optimizing checkpoint accessing performance. Based on the observation of data patterns in checkpoints, where the model weights change between adjacent checkpoints are minimal, making them suitable for delta compression. The proposed method implements delta compression across multiple interconnected training nodes and conducts experimental tests using real checkpoints generated during deep learning model training. The results demonstrate that, during the model training,delta compression has good compression effect for most checkpoints. Furthermore, the paper introduces dynamic intervals in delta compression to balance compression ratio and storage overhead, while also analyzing the characteristics of momentum datas. The analysis of existing methods and the optimization of checkpoint accessing performance offer insights for accelerating LLMs training.关键词：checkpoint;data compression;performance improvement97|44|0更新时间：2026-03-15 - 摘要：Accurate perception of disaster event situational relies on the timely and effective acquisition of relevant data carrying event information, as well as in-depth understanding and analysis of such data. Social media big data contains a wealth of event information. However, its characteristics of being voluminous, unstructured, and spatiotemporally sensitive pose significant challenges for the dynamic and complex awareness of disaster event situational. From the perspective of social media big data, we firstly constructed a causal knowledge graph of disaster events to effectively integrate heterogeneous information from social media big data, addressing the issues of unstructured data and spatiotemporal sensitivity. Secondly, we leveraged large language models and fine-tuning techniques to enhance the reasoning capability of disaster event evolution processes. Moreover, through the fine-tuned generative pre-trained model, we could more accurately identify causal sub-events of disaster events that were targeted and practical, effectivelyaddressing the challenges brought by the large volume of data and information redundancy. Finally, a disaster event awareness system was designed to assist relevant personnel in quickly and comprehensively understanding and analyzing disaster situations through user-system interaction. Experimental results show that the system achieves an average F1 score of 0.891 in disaster event-related text classification tasks, significantly outperforming baseline models. In terms of causal relationship generation, the fine-tuned generative pre-trained model can more accurately identify targeted and practical causal sub-events of disaster events, effectively improving the accuracy and efficiency of disaster event situational awareness.关键词：causal event;disaster information management system121|59|0更新时间：2026-03-15
- 摘要：Knowledge concept extraction has important application value in the fields of education, medical care, and finance. Knowledge concept extraction is a sub-task of named entity recognition. However, due to the lack of data sets and the particularity of knowledge concept entity types, directly applying general named entity recognition methods to knowledge concept extraction tasks often has poor results. In view of the above challenges, a method based on few-shot learning and chain-of-thought prompting for knowledge concept extraction was proposed, utilizing open-source large language models. Firstly, text representations focusing on entity semantics were trained through contrastive learning, and the relevance of the retrieved few-shot examples was enhanced using the K-nearest neighbors algorithm. Secondly, a method utilizing chain-of-thought prompting was adopted to present the samples, with the aim of improving the reasoning ability of large language models in knowledge concept extraction. Experimental results on multiple datasets demonstrate that the few-shot learning and chain-of-thought prompting for knowledge concept extraction method, onthe whole, has shown results superior over existing methods.关键词：named entity recognition;large language model79|55|0更新时间：2026-03-15
-
End-to-end seismic signals denoising via deep residual convolution and self-attention mechanisms AI导读
摘要：Denoising of seismic waveform signals is crucial for seismic monitoring and seismological research. To this end, we propose an end-to-end deep learning method for denoising seismic waveforms. The method combines the deep convolutional network with the multi-head self-attention mechanism. We employ a residual encoder-decoder structure, which is particularly well-suited for processing signals with complex backgrounds and rich details. At the same time,the multi-head self-attention mechanism can capture long-range dependencies. By jointly constraining the model with a consistent correlation loss and a frequency-domain mean squared error loss, outstanding denoising performance is achieved in both the time and frequency domains. Evaluation of the publicly available dataset STEAD shows that our method outperforms traditional and existing deep learning methods in two key metrics: peak signal-to-noise ratio (PSNR) and signal correlation coefficient (CC), achieving a Pearson correlation of 0.918 and a PSNR of 36.79, reaching the state-of-the-art performance. Furthermore, we have further validated our method using the seismic waveform data recorded by the Beijing-Tianjin-Hebei earthquake early warning network, which started operating in 2021, and the results indicate that our approach can effectively suppress noise while better preserving seismic signals (achieving a CC of 0.70 and a PSNR of 35.26).关键词：deep learning;convolutional network;self-attention mechanism;residual encoder-decoder91|143|0更新时间：2026-03-15 - 摘要：Currently, LLM-based agent technology is experiencing explosive evolution, and its development, collaboration, and full lifecycle management have become core issues in the industry. Based on industry practices, this article identifies typical application scenarios, conducts an in-depth analysis of the key technologies involved in agent development and communication, and proposes a comprehensive framework for the development and operation of agents. Finally, it explores the future development trends of agents.关键词：agent application;agent development platform326|90|0更新时间：2026-03-15
STUDY
-
Railway data management architecture and typical scenario application schemes based on data weaving AI导读
摘要：With the deepening of digital transformation in the railway industry, efficient management and integrated utilization of multi-source heterogeneous data have become key challenges for improving operational efficiency. This paper proposes a novel railway data management architecture based on data weaving technology. By constructing a seven-layer collaborative system-comprising the data source layer, metadata processing layer, data service layer, and four additional supporting layers-it enables cross-source integration and real-time sharing of data from business systems such as train–ground communication, mechanical operations, electrical supply, and rolling stock management. The study focuses on critical technologies, including proactive metadata management, knowledge-graph-driven data cataloging, and unified cross-source querying. It also designs a typical application scenario for collaborative fault handling in track and equipment maintenance. Experimental results demonstrate that the proposed architecture effectively breaks down data silos, improves fault resolution efficiency, and optimizes collaborative management. Thisarchitecture substantially enhances railway data governance capabilities and offers an extensible technical paradigm for the industry’s digital transformation.关键词：proactive metadata management;knowledge graph;cross-source querying;collaborative fault handling110|70|0更新时间：2026-03-15
APPLICATION
-
Exploring data governance pathways for grassroots government in the context of trusted data space AI导读
摘要：Against the backdrop of advancing national data strategies and the construction of digital government, trusted data space, as the institutional vehicle for achieving high-quality data governance, is increasingly becoming an important foundation for the development and utilization of government data. This study was centered on the collaborative theoretical mechanisms and the right-responsibility structure within trusted data space, and legal challenges faced by government data governance were analyzed，that included the divergence between new and traditional governance paradigms, tensions between institutional design and operational practice, and the imbalance of authority and responsibility that had resulted in phenomena such as “data suspension.”In response, a three-dimensional framework comprising the legal guidance layer, collaboration layer, and supervision layer was proposed to reconstruct grassroots government data governance. This framework was intended to reshape the development model of trusted data spaceand to promote the professionalization and legalization of data governance. Ultimately, it enhances the data governance capacity of grassroots government and realizes a smart and intelligent digital government.关键词：government data;trusted data space;data governance paradigm85|84|0更新时间：2026-03-15
FORUM
- Postal code：100079
- Tel：（010）53879208、53859533 Email：bdr@bjxintong.com.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：273110 Visits Today：16
0