Under the background of policy guidance and technological development, the Multi-RevRobot for the translation of clinical evidence of traditional Chinese medicine (TCM) was developed to enhance the efficiency and applicability of clinical evidence translation in TCM. A prototype of Multi-RevRobot was designed and developed. This involved components such as the perception system, control system, execution system, power system, and interaction system. Universal model interfaces were integrated to connect with vertical large models and knowledge bases, enabling intelligent interaction. The artificial potential field method was combined with reinforcement learning to achieve efficient coordination between the robot's global path planning and local obstacle-avoidance decision-making. This allows the robot to be deployed and applied in various scenarios. Multi-RevRobot features intelligent interaction, report generation, and multi-scenario adaptability. It provides a practical solution for TCM clinical evidence translation and plays corresponding roles in various scenarios. Multi-RevRobot is expected to serve as a clinical and research assistant for doctors, researchers, patients, and other groups. It can improve the efficiency of TCM-related work and facilitate the intelligent and efficient translation and application of TCM evidence.

Smoking remains a leading cause of preventable morbidity and mortality worldwide, and short-video platforms have increasingly become influential channels for disseminating smoking cessation–related information to the public. TikTok and Bilibili are two major short-video platforms with large user bases and distinct characteristics, exerting considerable influence on health perceptions, particularly among younger populations. However, due to their open nature and limited quality control, the quality and reliability of smoking cessation–related content vary substantially, with some videos containing inaccurate or misleading information, posing a public health challenge. This study aimed to systematically evaluate the quality and usefulness of smoking cessation–related videos on TikTok and Bilibili and to assess whether these platforms serve as valuable resources for smokers attempting to quit. In January, 2024, a search was conducted on TikTok and Bilibili using the keywords “quit smoking (戒烟)” and “how to quit smoking yourself (如何自己戒烟).” Four independent reviewers assessed 400 videos using a standardized coding framework. Video quality and content were evaluated using global quality score (GQS), Usefulness Score (US) and Message Appeals (MA). TikTok videos showed higher levels of user engagement than those on Bilibili; however, no statistically significant differences were observed between the two platforms in Global Quality Score (GQS) or Usefulness Score (US). Most videos received a GQS of 3 (Bilibili: 56.5%; TikTok: 55.5%) and were rated as slightly useful, with fear appeals being the most commonly used message strategy. Videos produced by health professionals and institutions demonstrated higher quality and usefulness. Correlation analyses revealed distinct platform-specific patterns: on TikTok, video duration and time-normalized engagement indicators (daily video likes, daily video comments, daily video saves, and daily video shares) were positively associated with GQS, US, and message appeal, whereas on Bilibili, engagement indicators showed weak or negative associations with video quality. There remains substantial room for improvement in the overall quality of smoking cessation–related short videos on both platforms. Greater involvement of health and communication professionals, as well as relevant institutions, may help enhance content quality. Achieving an appropriate balance between scientific accuracy and public appeal is essential for maximizing the effectiveness of short videos as smoking cessation health education tools.

ABSTRACT We present Multi-Epoch Spectroscopic Solver (MESS), a fully automated algorithm for identifying and characterizing double-lined spectroscopic binaries ($\mathcal {SB}2$) in large data bases of multi-epoch spectra. MESS extends the two-dimensional TODCOR approach to a global multi-epoch formalism, deriving the radial velocities (RVs) of both components at each epoch while optimizing the templates jointly across all observations. Template optimization searches a continuous synthetic-spectra manifold spanning an eight-dimensional parameter space: effective temperature, surface gravity, and rotational broadening for each star, together with a common metallicity and the flux ratio. Single-lined spectroscopic binaries ($\mathcal {SB}1$) and single stars ($\mathcal {S}1$) are handled within the same framework by fitting one optimized template, with either epoch-dependent RVs ($\mathcal {SB}1$) or a single shared RV ($\mathcal {S}1$). Model selection among $\mathcal {S}1/\mathcal {SB}1/\mathcal {SB}2$ uses the Bayesian information criterion with an effective sample size that accounts for intra-spectrum correlations, and is complemented by the Wilson relation between the two RVs to infer the mass ratio and systemic velocity without a full orbital solution. We validate MESS on 1500 simulated LAMOST MRS systems (signal-to-noise ratio $=50$), with primary RV semi-amplitudes predominantly below the instrumental resolution, achieving an overall classification accuracy of $\sim 95~{{\ \rm per\ cent}}$. We also derive full orbital solutions for two $\mathcal {SB}2$ systems detected in our LAMOST analysis, including a faint-secondary case with flux ratio $\sim 0.1$, and present example outputs for one $\mathcal {SB}1$ and three constant-velocity stars. A companion paper will report the survey-wide application to LAMOST DR11 and the resulting $\mathcal {SB}1/\mathcal {SB}2$ catalogues.

Short videos have become increasingly popular in recent years, leading to the rise of many short video platforms with large user bases, such as leading platforms like TikTok. The user groups on these platforms are vast. With the rise of short-video platforms, TikTok, as one of the leading players, has become a highly influential social media platform worldwide. With the development of short videos, personalized recommendations are also playing an increasingly important role in short videos and are constantly evolving. This report examines the definition, classification, and evolution of personalized recommendation systems, along with their applications and the development trends and suggested countermeasures in short video platforms, with the objective of analyzing the current status and emerging trends of personalized recommendations in this domain. Research has found that personalized recommendations are of great help to the development of short videos. Enterprises need to keep up with the times and constantly update their personalized recommendation systems to enhance user experience and retain more users. Meanwhile, users also need to improve their ability to distinguish during the process of using short videos. In addition, the government should supervise enterprises to ensure the healthy application and development of personalized recommendations in short videos. It provides valuable reference information for content creators, short video platforms, and end users.

As an emerging e-commerce model, live-streaming commerce has attracted widespread attention from merchants. Since many merchants do not have their own anchors, they need to hire professional anchors, who are called commissioned live-streaming commerce. Merchants can choose either agent anchors or celebrity anchors, that is, a choice between two different live-streaming channels. Choosing celebrity anchors with large fan bases can crank up higher sales for merchants, but comes with higher commission expenses and potential competition from the anchors’ own brand products. Conversely, choosing agent anchors whose popularity is far less than that of celebrity anchors may still bring high profits, as these agents charge lower commissions. Inspired by this interesting phenomenon, the present study examines the challenges faced by merchants in live-streaming channel selection using commissioned live-streaming commerce. The study employs a Cournot model to investigate the issues at hand and analyze the impact of different factors on merchants' channel selection based on simulation. The results show that when the effort cost of a merchant choosing anchors is high and the merchant's brand image advantage is significant, the merchant is more inclined to choose agent anchors; otherwise, the inclination is to choose celebrity anchors. The findings provide support for the decision-making process of different parties involved in live-streaming commerce and thus, promote the sustainable development of the live-streaming commerce industry.

The integration of different matrix retrieval technologies is likely to shape the future direction of technological development. By combining traditional methods such as TF-IDF with the powerful ability of deep learning to process high-dimensional data, which can build precise retrieval systems. These systems integrate graph neural networks with emerging tools effectively, such as PageRank, improving link analysis by directly combining node attributes like entity types and content details into the sorting process, therefore enhancing context understanding capabilities. For the efficient execution of complex computations similar to transformer models, innovative strategies like scarce attention are crucial techniques, such as sliding window attention and low-rank approximation play a crucial role in handling long text, large code bases, and multiple iterative searches. In addition, technological advancements in dedicated hardware devices like TPU have lessened the challenges brought by intensive matrix computations, making advanced real-time search tasks that were previously unachievable possible. The coordinated development of algorithms and hardware is elevating the semantic parsing capabilities of search systems to a remarkable height. Future search applications will deeply integrate context and personalized cognition, establishing benchmark standards for a comprehensive understanding of complex data, thereby changing educational models and information acquisition methods.

Background A scientific understanding of soil erosion in coal mining subsidence areas is crucial for promoting green development and ecological restoration within the large coal bases of the Yellow River’s middle reaches. Methods This study focused on the shallow soil layer (0–40 cm) of a loess mining subsidence area and typical slopes in northern Shaanxi. Integrating field sampling, laboratory measurements, and model-based calculations, we analyzed the impact of mining subsidence on soil anti-erodibility by examining the coupled “slope position × soil depth” effect. Results The results indicate that: (1) Mining subsidence significantly reduced the content of water-stable aggregates (>0.25 mm) and decreased the mean weight diameter of soil aggregates on the loess slope, with the >0.25 mm aggregate content being the most affected; (2) The adverse effect of subsidence on key soil anti-erodibility indicators weakened progressively from the slope top to the slope toe. The 0–20 cm soil layer at the slope top showed the greatest sensitivity; (3) Following subsidence, the comprehensive soil anti-erodibility index decreased significantly, indicating markedly enhanced erosion susceptibility. The most pronounced reduction (97.13%) was observed in the “slope top + 0–20 cm soil layer”. Conclusion The “slope top + 0–20 cm soil layer” should be prioritized as a key area for erosion control on subsided loess slopes. These findings provide a scientific basis for targeted soil erosion management in coal mining subsidence areas of northern Shaanxi.

Mahanadi Coalfields Limited (MCL), a subsidiary of Coal India Limited, plays a vital role in meeting India's energy needs by ensuring a steady supply of coal to power plants and core industries.As a major public sector enterprise, the company operates in a highly regulated environment where transparency, financial discipline, and strict adherence to tax laws are essential.In this context, compliance with Goods and Services Tax (GST) and Tax Deducted at Source (TDS) becomes central to how MCL manages its day-to-day business and long-term sustainability.This study explores how GST and TDS compliance actually works inside MCL, moving beyond regulations on paper to the practical routines followed by departments such as finance, accounts, and procurement.It looks at how bills are processed, how GST returns are filed, how input tax credit is tracked, and how TDS is deducted and deposited on time, as well as the communication and coordination required between different units.The research also pays attention to the difficulties employees face-such as frequent changes in tax rules, complex documentation, and system-related issues-and how these challenges affect cash flow, reporting accuracy, and internal control.By drawing insights from company records, employee experiences, and relevant legal provisions, the study aims to present a grounded picture of MCL's GST and TDS practices rather than a purely theoretical view.The findings are expected to highlight both the strengths of MCL's present system-such as standardised procedures and audit mechanisms-and the areas where further improvements, training, or technology support could make compliance smoother and more efficient.Ultimately, the study hopes to offer practical suggestions that can help MCL and similar public sector organizations strengthen their tax compliance culture and support better financial governance. OBJECTIVES OF THE STUDYThe study on "GST and TDS Compliance Practices: A Case of Mahanadi Coalfields Ltd (MCL)" is undertaken with the following key objectives Primary Objective 1.To understand how MCL handles GST.This includes invoicing, return filing, and input tax credit.2. To study how MCL manages TDS.This covers deduction, deposit, and reporting of TDS.3. To check whether MCL's internal processes help in timely and correct tax compliance. Secondary Objectives1. To find out the major challenges MCL faces in GST and TDS compliance.2. To see how GST and TDS practices impact MCL's daily work and financial performance.3. To compare MCL's practices with general PSU standards.4. To suggest simple improvements that can make MCL's GST and TDS compliance better. REVIEW OF LITERATURE Kumar (2018) -Discussed the impact of the Goods and Services Tax on large Indian companies and reported that the shift to a multi rate GST structure increased the need for accurate classification, timely return filing, and strict documentation, especially in capital-intensive sectors such as mining. Sharma and Gupta(2019)-Examined GST compliance in public sector enterprises and found that organisations with strong internal control and regular employee training were better able to avoid interest, penalties, and mismatch of input tax credit. Rao (2020)-Studied GST implementation issues in the coal and mining industry and highlighted challenges related to valuation of services, place of supply rules, and reconciliation of invoices raised by numerous contractors and transporters. Mehta and Verma (2020)-Analysed TDS practices in large corporations and showed that errors in deduction and late deposit of tax were mainly due to frequent changes in TDS provisions and lack of integrated accounting software. Singh (2021)-Focused on the role of technology in GST and TDS compliance and concluded that ERP systems, e -invoicing, and automated reconciliation significantly reduced manual mistakes and improved transparency in tax reporting. Kaur and Joshi (2022) -Compared tax compliance in public and private companies and observed that while public sector units had well-defined procedures, they were slower in adapting to amendments in GST and income-tax laws, which sometimes led to operational delays. Patel (2023) -Presented a case study on a state owned mining company and found that decentralised operations, large vendor bases, and complex contract structures made GST and TDS compliance difficult, underscoring the need for central monitoring and periodic review of tax processes. Research Objective1. To understand how Mahanadi Coal Field Ltd. manages its GST compliance. 1.To explore how the company handles TDS compliance in day-to-day operations. 2.To identify the main challenges and hurdles the company faces in GST and TDS processes.

BackgroundAs a high tuberculosis (TB) burden area in China, Dazu District of Chongqing Municipality contains a large rural population and exhibits typical features of TB endemicity.ObjectiveThis study aimed to analyze the epidemiological characteristics and treatment outcomes of pulmonary tuberculosis (PTB) in Dazu District from 2005 to 2024, with the aim of supporting the optimization of regional TB control strategies.MethodsData on PTB cases in Dazu District from 2005 to 2024 were collected from the China Disease Control and Prevention Information System. Descriptive epidemiological methods were employed to analyze the temporal, demographic, and geographical distributions, along with trends in treatment outcomes. Global and local spatial autocorrelation analyses were performed using Moran I and Getis-Ord Gi* statistics, respectively.ResultsA total of 10,236 cases were reported, for an average annual notification incidence of 65.2 per 100,000 population. The annual average notification incidence decline rate was 7.7%. Joinpoint regression analysis revealed a statistically significant decline in annual incidence rates (average annual percent change=−6.81, 95% CI −7.25 to −6.30, P<.0001). The bacteriological positivity rate initially decreased before rising, reaching 81.6% in 2024. Reported case counts peaked in March, while relatively lower numbers were observed during October, November, and December. Cases were predominantly among male patients, with a male-to-female ratio of 3.57:1. The case composition ratio in the ≥65 years age group has gradually increased, from 13.8% in 2006 to 19.9% in 2015 and to 38.5% in 2024. Occupational distribution was primarily among farmers (77.6%, 7948/10,236), homemakers or unemployed individuals (5.6%, 570/10,236), and students (3%, 303/10,236). Cases were concentrated in Longshui Town (13.1%, 1339/10,236), Tangxiang Subdistrict (9.1%, 933/10,236), and Longgang Subdistrict (7.9%, 811/10,236)—areas with large population bases. Among these, Guoliang Town exhibited the highest average annual notification incidence (314.4/100,000). Treatment success rate reached 91.3%. Multivariate binary logistic regression revealed that age 25‐44 years (OR 1.755; 95% CI 1.320‐2.332; P<.0001), undergoing initial treatment (OR 3.786; 95% CI 2.524‐5.680; P<.0001), absence of HIV coinfection (OR 2.499; 95% CI 1.714‐3.643; P<.0001), negative bacteriologic test results (OR 2.841; 95% CI 2.214‐3.646; P<.0001), and the receipt of full-course supervised treatment (OR 7.705; 95% CI 4.520‐13.137; P<.0001) were significantly associated with treatment success.ConclusionsThe notification incidence of PTB in Dazu District, Chongqing, has gradually declined. Particular focus is required on the treatment of young children, elderly individuals, patients with HIV coinfection, those under intensive phase supervision, bacteriologically positive cases, and retreatment cases. These measures may reduce the incidence of PTB and improve treatment success rates in our district.

The rapid spread of fake news across social media poses significant threats to politics, economics, and public health. During the COVID-19 pandemic, social media influencers played a decisive role in amplifying misinformation due to their large follower bases and perceived authority. This study proposes a Multi-Stage Detection System for Influencer Fake News (MSDSI-FND) to detect misinformation propagated by influential users on the X platform (formerly Twitter). A manually labeled dataset was constructed, comprising 68 root tweets (42 fake and 26 real) and over 40,000 engagements (26,700 replies and 14,000 retweets) collected between December 2019 and December 2022. The MSDSI-FND model employs a two-stage analytical framework integrating: (1) content-based linguistic and psycholinguistic analysis, (2) user profiles analysis, structural and propagation-based modeling of information cascades analysis. Several machine-learning classifiers were tested under single-stage, two-stage, and full multi-stage configurations. An ablation study demonstrated that performance improved progressively with each added analytical stage. The full MSDSI-FND model achieved the highest accuracy, F1-score, and AUC, confirming the effectiveness of hierarchical, stage-wise integration. The results highlight the superiority of the proposed multi-stage, influential user-aware framework over conventional hybrid or text-only models. By sequentially combining linguistic, behavioral, and structural cues, MSDSI-FND provides an interpretable and robust approach to identifying large-scale misinformation dissemination within influential user-driven social networks.

Implementing distributed cloud-based applications commonly at the basis of user-facing services goes through several challenges. In particular, such applications must be scalable to accommodate increasingly large user bases, providing consistency on accesses to shared data while executing on highly distributed concurrent commodity hardware. In addition, as these applications are subject to workload fluctuations, they must be elastic , i.e., able to scale out to accommodate workload increases as well as to scale back in to avoid over-provisioning and thus unnecessarily high costs in case of workload decreases. This paper presents AEON, a programming framework that supports the development of scalable elastic cloud-based distributed applications. In short AEON leverages two synergistic “levels” of programming: I. An application programming language (APL) allows programmers to conceive scalable applications using the popular actor paradigm, augmented with an intuitive notion of event to capture non-interleaved executions across multiple actors as needed for non-trivial shared data, all the while avoiding error-prone manual concurrency control. That is, based on a simple type-based ownership analysis asserting that references in AEON applications follow a DAG-based referencing structure, events are executed efficiently in a serializable fashion leveraging a lightweight synchronization protocol which is also exploited for creating consistent snapshots of the distributed application’s shared data. II. An elasticity programming language (EPL) allows application managers to define policies for guiding efficient fine-grained automated scaling — in and out — of applications at runtime. While these policies refer to applications written with I, they only refer to high-level abstractions in those (e.g., types of actors and methods), are inversely not referred to by them, and avoid side-effects to minimize effects on application performance. After presenting our programming framework with its language design choices and runtime system implementation, we present a study applying it to several use cases, and evaluate its performance. In short, our APL’s synchronization model scales better than manual locking or the use of automated traditional two-phase locking with existing actor languages, or the use of an external transactional store; under workload fluctuations our EPL allows programs to be executed with significantly improved performance without increased resource usage, or with similar performance but significantly fewer resources.

Background The structure of scientific collaboration networks has been recently studied in different disciplines. The analysis of large data bases using specialized software capable of connecting the different authors and coauthors in a large network of scientific collaboration has enabled the construction of collaboration graphs between investigators in several disciplines. The use of these networks is not new in the field of bibliometrics; however, attention has been recently focused on academic co-authorship networks which might be more representative of the structure of knowledge of an academic community. Objective To describe the structure of scientific collaboration networks in Argentina based on co-authorship network analysis of the articles published in the field of cardiology during 2007. Material and Methods We conducted a bibliographic search of Argentine papers published in the field of cardiology. Data was retrieved from Medline and from two local journals: Rev Argent Cardiol and Rev Fed Arg Cardiol. Collaboration networks between authors were constructed using the Kamada-Kawai algorithm included in the Pajek software. Results Mean papers per author ranged from 1.12 to 1.24, the exponent tau of productivity was 2.78 to 3.45, mean authors per paper from 3.60 to 6.51, and collaborators per author ranged between 2.60 and 4.88. The construction of collaboration networks showed that the size of the giant component was between 13.1% and 65.8%, the mean distance between authors was 1.5 to 8.5 and the maximum distance was 5-24. Conclusions The structures of different scientific collaboration networks based on co-authorship in Argentine papers published in local and international journals were studied. The productivity index followed Lotka.s law with a value that was similar to the one reported in biomedical publications. The size of the collaboration network was smaller than expected, probably due to the short period of the study. The mean distance between authors was greater than we expected, indicating an inadequate structure of connections and collaboration between investigators.

As technology continues to advance, it becomes increasingly integrated into daily life facilitating complex tasks across a range of environments. While some applications such as smartphones and smartwatches are less critical, others like healthcare devices and autonomous vehicles demand bug-free performance to prevent financial loss or harm. Traditionally, simulation-based testing and formal verification played a major role in ensuring a bug-free device. However, the simulation of bigger systems is limited to a definite number of scenarios on the Design under Verification (DUV). Hence, it is unable to explore all possible inputs that can occur. Formal verification, on the other hand, offers a higher level of assurance through mathematical proofs but is both time-consuming and suffers from scalability issues, especially as designs grow in complexity. Recently, Large Language Models (LLMs) have shown promise in tasks previously limited to human expertise. Their natural language processing capabilities can assist in handling extensive specifications and source code, particularly in debugging hardware descriptions and analyzing security and functionality. The utilization of Retrieval Augmented Generation (RAG) has further enhanced LLMs by incorporating large specification or source code bases, thereby improving their bug-identification and correction capabilities. While recent advancements in LLMs, particularly with RAG, have yielded promising results in bug identification and correction for a small class of hardware bugs, significant gaps remain in their full potential for systematically addressing a wide range of hardware bugs. For instance, existing LLM methodologies struggle to detect bugs involving incorrect constant values, i.e., the use of wrong constants in source code. This limitation underscores the need for further exploration in utilizing LLMs to fully optimize the verification process. To bridge this gap, we propose a 3-phased 4-stage LLM-assisted systematic bug closure methodology that focuses on functional bugs in Verilog HDL rather than structural or syntactic issues. Our approach extracts functional properties of the DUV and systematically breaks down complex expressions into smaller sub-expressions to facilitate bug detection and correction. By employing RAG, the LLM is guided using the functional specifications and source code to identify and correct bugs. If the initial guidance through RAG is insufficient, our methodology initiates an iterative bug closure process. This includes incorporating more extensive information from the specifications, fetching additional lines of code for bug localization, and breaking down complex Verilog HDL expressions. In our comprehensive evaluation, we assess the LLM’s capabilities using 9 different categories of bugs. As benchmarks, we use 5 OpenTitan Intellectual Property (IP) cores to demonstrate the scalability and effectiveness of our bug closure methodology where ≈ 60% of the bugs were corrected. Specifically, we evaluate OpenAI’s GPT-4 in its ability to identify and correct functional bugs in Verilog HDL code.

Social media platforms enhance the propagation of online misinformation by providing large user bases with a quick means to share content. One way to disrupt the rapid dissemination of misinformation at scale is through warning tags, which label content as potentially false or misleading. However, past warning tag mitigation studies yield mixed results for diverse audiences. We hypothesize that personalizing warning tags to the individual characteristics of their diverse users may enhance mitigation effectiveness. To reach the goal of personalization, we need to understand how people differ and how those differences predict a person's attitudes and behaviors toward tags and tagged content. In this study, we leverage Amazon Mechanical Turk (n = 132) and undergraduate students (n = 112) to provide this foundational understanding. With all participants combined, we find attitudes towards warning tags and self-described behaviors are significantly influenced by factors such as Need for Cognitive Closure (NFCC), Political orientation, and Trust in Medical Scientists when controlled for covariates such as age and recruiting platform. Analyses of each sample further show that tag attitudes were influenced by Trust in Religious Leaders, and Big Five Inventory (BFI) traits for Openness and Conscientiousness. We synthesize these results into design insights and a future research agenda for more effective and personalized warning tags and misinformation mitigation strategies more generally.

Earth System Models (ESMs) are critical for understanding past climates and projecting future scenarios.However, the complexity of these models, which include large code bases, a wide community of developers, and diverse computational platforms, poses significant challenges for software quality assurance.The increasing adoption of GPUs and heterogeneous architectures further complicates verification efforts.Traditional verification methods often rely on bitwise reproducibility, which is not always feasible, particularly under new compilers or hardware.Manual expert evaluation, on the other hand, is subjective and time-consuming.Formal methods offer a mathematically rigorous alternative, yet their application in ESM development has been limited due to the lack of climate model-specific representations and tools.Here, we advocate for the broader adoption of formal methods in climate modeling.In particular, we identify key aspects of ESMs that are well suited to formal specification and introduce abstraction approaches for a tailored framework.To demonstrate this approach, we present a case study using CIVL model checker to formally verify a bug fix in an ocean mixing parameterization scheme.Our goal is to develop accessible, domain-specific formal tools that enhance model confidence and support more efficient and reliable ESM development.

Modern neural network methods combine work with an axiomatic mathematical description (laws, equations, invariants, logical rules) and the power of neural networks for learning from data, pattern recognition and differentiation through complex spaces. This combination produces systems that can learn from data, observe given laws and, as a result, make predictions, solve problems and even discover new hypotheses. Quality depends on the formulation of axioms and the presence of correct formulations, the complexity of scaling to very large axiomatic bases, trade-offs between the accuracy of fitting to data and compliance with laws, interpretation and verification of results. Modern neural network methods with an axiomatic mathematical description have better generalization and physical interpretability due to compliance with axioms, the ability to work with small data due to built-in laws and the ability to discover new dependencies within the framework of formalized rules. Theoretical principles and formal axioms set requirements for neural networks and their training so that solutions to scientific problems correspond to the laws of nature, invariances, data characteristics and other desired properties. Power: an axiomatic neural network tends to be accurately modeled given its sufficient complexity and large scientific data and knowledge. The author proposes a neural network axiomatic solver coaching AGI method for solving scientific and practical problems according to their formulations and developed systems of axioms.

Modern neural network methods combine work with an axiomatic mathematical description (laws, equations, invariants, logical rules) and the power of neural networks for learning from data, pattern recognition and differentiation through complex spaces. This combination produces systems that can learn from data, observe given laws and, as a result, make predictions, solve problems and even discover new hypotheses. Quality depends on the formulation of axioms and the presence of correct formulations, the complexity of scaling to very large axiomatic bases, trade-offs between the accuracy of fitting to data and compliance with laws, interpretation and verification of results. Modern neural network methods with an axiomatic mathematical description have better generalization and physical interpretability due to compliance with axioms, the ability to work with small data due to built-in laws and the ability to discover new dependencies within the framework of formalized rules. Theoretical principles and formal axioms set requirements for neural networks and their training so that solutions to scientific problems correspond to the laws of nature, invariances, data characteristics and other desired properties. Power: an axiomatic neural network tends to be accurately modeled given its sufficient complexity and large scientific data and knowledge. The author proposes a neural network axiomatic AGI method for solving scientific problems according to their formulations and developed systems of axioms.

ABSTRACT The field of recommendation systems is a hot topic thanks to the increasing number of available digital products and services. In connection with this topic, the research of Graph Neural Network solutions has played a significant role in recent years. Research and development of an online recommendation system that also manages the challenges of a rapidly changing environment are important from a practical point of view as well. Our aim was to develop an approach that possesses scalable inference and adaptation and uses latent features. The main contribution of this paper is the development of a candidate generation process for online collaborative filtering on implicit feedback data that can scale to large user and item bases. We proposed multiple ways how embeddings can be obtained in a fast and scalable way, namely Lookup, Inductive neighbor aggregation, Neighbor aggregation with importance scores, and GraphSAGE‐based Graph Neural Network (GraphSAGE+) method with continuous representation update for online learning. By combining these inductive and transductive methods for the embeddings, we developed a novel online Collaborative Filtering approach. We evaluated our approach on two e‐commerce datasets and found that it outperformed traditional recommendation algorithms such as Matrix Factorization.

Complementors are critical stakeholders in two-sided digital platforms, but their continued adoption intentions of these platforms remain underexplored compared with those of users. This study extends the Expectation-Confirmation Model (ECM) to address this gap by substituting perceived usefulness with perceived value-a construct that captures complementors' holistic cost-benefit evaluations-and examining how same-side network effects, cross-side network effects, perceived compatibility, and perceived complementarity influence perceived value, satisfaction, and continued adoption intentions. Using survey data from 285 complementors (third-party developers) in the Android ecosystem, we tested our hypotheses via PLS-SEM. The results show that (1) same-side network effects (intense competition among complementors) negatively impact perceived value; (2) cross-side network effects (large user bases), perceived compatibility (technical alignment with the platform), and perceived complementarity (ecosystem resource support) positively influence perceived value; (3) perceived value increases both satisfaction and continued adoption intentions; and (4) satisfaction further drives continued adoption intentions. Fuzzy set qualitative comparative analysis (fsQCA) identifies four distinct configurational paths to complementors' continued adoption intentions, highlighting that these intentions arise from the interplay of multiple factors (e.g., low same-side competition combined with high compatibility, or strong cross-side effects paired with high complementarity). This study advances the theoretical understanding of complementor behavior by extending the ECM to platform contexts, and it offers actionable insights for managers seeking to retain complementors through balancing competition, enhancing the user scale, and improving technical and ecosystem support.

Enterprise cloud systems face increasing vulnerability to sophisticated cyberattacks that exploit conventional authentication mechanisms relying on static credentials and basic multi-factor authentication approaches. This article presents an adaptive authentication framework that leverages behavioral biometrics to dynamically adjust security requirements based on real-time risk assessments, addressing the critical gap between robust security and seamless user experience in cloud-based enterprise environments. The proposed article captures distinctive behavioral patterns, including keystroke dynamics, mouse movement characteristics, and application interaction sequences, to establish unique user baselines that are difficult for attackers to replicate. A comprehensive risk assessment engine integrates these behavioral signals with contextual information such as device trust scores, geolocation analysis, and temporal access patterns to trigger appropriate authentication responses only when security risks warrant additional verification. Through controlled testing in simulated enterprise cloud environments, the adaptive authentication system demonstrated substantial improvements in detecting and preventing account compromise attempts while minimizing disruption to legitimate user workflows. The behavioral biometric approach proved particularly effective against sophisticated attacks, including credential stuffing, social engineering, and insider threats that frequently bypass traditional security controls. User experience evaluations revealed high acceptance rates and minimal productivity impact when risk thresholds were appropriately calibrated, indicating a successful balance between security enhancement and operational efficiency. The modular system architecture facilitates integration with existing enterprise identity management infrastructure while supporting scalable deployment across large user bases. This article contributes to the advancement of zero-trust security principles by providing continuous verification capabilities that strengthen enterprise cloud security postures without requiring complete infrastructure overhauls, offering organizations a practical pathway toward more resilient authentication strategies in an evolving threat landscape.