This article discusses modern methods of extrapolating pre-trained transformers aimed at improving their ability to process long and short text sequences in Russian in the financial sector. Particular attention is paid to the task of classifying texts that reflect broker analysts’ expectations regarding market movements (expectations of growth, decline, or uncertainty of change). To solve this problem, the application of lightweight language models ruBERTtiny1 and ruBERT-tiny2 is investigated, which are adapted to work effectively with large amounts of input data while maintaining prediction quality. The paper analyzes various approaches to expanding the contextual window of models, including extrapolation methods, and considers the impact of tokenization, vectorization, and embedding strategies on the final classification results. Additionally, the paper discusses the peculiarities of using transformers in conditions of increased market volatility and changing news flows, which allows for a more in-depth assessment of the stability of the proposed solutions. Furthermore, a formula for calculating a leading indicator for stock markets is proposed and discussed, demonstrating the practical significance of using transformer models in the analysis of financial texts and the formation of analytical metrics. The presented results highlight the promising application of compact transformers in predictive financial analytics tasks.

This paper explores the application of machine learning methods for sentiment analysis of user-generated texts in the Russian social network VKontakte. The sentiments of millions of users could be monitored and analyzed in real time, that facilitates prompt decision making and forecasting of social processes. Textual data, including posts and comments, were collected via the VK API. The preprocessing pipeline involved text cleaning, lemmatization, stop-word removal, and TFIDF vectorization. Several classification models were tested, including logistic regression, random forest, and naïve Bayes, as well as deep learning models such as LSTM and Transformers (RuBERT). The naïve Bayes classifier demonstrated the best performance in terms of recall and overall metric balance. Sentiment analysis results revealed that the majority of user texts were neutral or positive, with only a small portion being negative. The paper includes visualizations and statistical summaries of sentiment distribution. The study confirms the effectiveness of classical machine learning methods for processing and analyzing textual data in Russian social networks.

The paper analyzes evolution and state-of-art of adventure videogames. The goal of the study is to reveal the trends in creating text quests based on innovative approaches. These are the tasks to solve: to analyze interactive fiction videogames, to make a review of the technologies employed in videogames development, to highlight social and cultural effects. The scientific value is the evaluation of the potential of adventure games in present-day conditions. The practical applicability is proved by real examples demonstrating the AI-tools used for creating elements of text quests. For example, neural networks are employed both to generate the plot of the quests and audio and visual content. That shortens development time and reduces the cost of production process. The comparative analysis performed for the game “Far Cry” allows revealing the ways for further improvements in the development process. The authors also consider the situation concerning blocking foreign online-platforms in Russia due to the disallowed content and map out the ways for domestic videogame market. The outcome of the study emphasizes the importance of government support for establishing national game-platform to reinforce state values and provide the protection of national culture.

The attention mechanism is a key component of modern artificial neural networks designed to process data of various nature. The article examines the dynamic of attention using a continuous model. In this model, attention is described as the movement of interacting tokens. It is shown that, under suitable assumptions, attention is Lipschitz continuous. In particular, Lipschitz continuity may be ensured by token normalization. The dynamics of transformers is modelled by a system of differential equations. Lipschitz continuity guarantees that there exists a solution to this system. The purpose of the study is to investigate the behavior of tokens that make up promt under an unlimited increasing in the number of transformer layers. For one-dimensional tokens, a qualitative description of the trajectories of tokens and the dynamics of the attention matrix is given. It is shown that if a token goes beyond a fairly narrow corridor at some point (the width is on the order of the logarithm of the promt size), this token tends to infinity (positive or negative, depending on which border the exit occurred). The research methodology is based on continuous parameterization of the attention matrix. The common representation of transformer dynamics by difference equations has been replaced by a representation using systems of ordinary differential equations. A huge number of publications are devoted to the description and study of transformers, but most of them do not contain accurate mathematical descriptions of architecture. This article attempts to give a mathematically meaningful and at the same time fairly simple description of attention. The description dynamics of 1-d tokens is certainly much simpler than the dynamics of multidimensional tokens. Nevertheless, this description gives an idea of the behavior of transformers in a more general situation creates a framework for future investigation.

Superplastic forming is an advanced technology used in the aerospace and automotive industries, as well as in the medical sector, for fabricating complex seamless components. However, its application is limited by high costs and the extended duration of the process. While finite element analysis in CAE systems such as ANSYS provides accurate results, it is computationally expensive. While finite element analysis performed in CAE systems such as ANSYS provides high-fidelity results, its computational expense creates a need for fast and accurate predictive models capable of supplementing or replacing this approach in multi-criteria analysis tasks. Despite the increasing adoption of machine learning across various disciplines, the development of reliable predictive models for specific geometric characteristics of superplastically formed components remains an understudied research area. The purpose of this study is to develop and verify a Gaussian process based model for predicting key geometric parameters of a hemisphere during the superplastic forming. An additional objective was to create an initial dataset using data generated from numerical simulations. The Latin Hypercube Sampling method was employed to design the experiment and generate the initial dataset, enabling efficient variation of material parameters K, m and pressure regime within ranges typical for aluminum alloys. Based on data from 50 numerical simulations, a predictive model for the hemisphere’s geometric characteristics was developed with Gaussian Process Regression with a composite kernel. Model hyperparameter optimization was performed using RandomizedSearchCV. The developed Gaussian Process Regression model demonstrated high accuracy, achieving a coefficient of determination greater than 0.90 on the validation set for all target variables: thickness at the pole, average height, and height difference. Analysis of the Mean Squared Error confirmed the models generalization capability and absence of overfitting. This research is aimed at integrating the model into a digital twin system for real-time optimization of process parameters. The main challenge in scaling this approach is the computational cost associated with generating the required training data.

This article presents a comprehensive analysis of the issue of ensuring digital safety for minors in the context of the rapid development of the information society and digital technologies. The objective of the study is to identify the legal, organizational, and technological mechanisms for protecting children from informational threats emerging in the online environment, including harmful content, cyberbullying, involvement in illegal activities, and manipulative algorithms of social media platforms. Special attention is given to the analysis of legal regulations governing this area, the institutional role of state authorities, and the importance of educational initiatives. The methodological basis of the study includes a systematic and interdisciplinary approach, taking into account the legal, sociocultural, pedagogical, and technological dimensions of the problem. The research employs methods such as content analysis of legal documents, comparative legal analysis, as well as synthesis of practical experience and sociological data. The key findings of the study highlight the fragmented nature of current legislation and institutional mechanisms in the field of digital safety for minors. The study substantiates the need for the development of a comprehensive multi-level protection model. This model includes the establishment of a specialized supervisory body, strengthening cooperation between schools, families, and children, and implementing modern technological tools for filtering and monitoring digital content. The conclusions of the article emphasize the importance of an integrated approach to addressing the identified problem, the need for coordinated interagency cooperation, the promotion of digital literacy among all participants in the educational process, and the improvement of legal frameworks. The recommendations and initiatives presented in this study may serve as a foundation for the development of public policy in the area of digital safety, the design of educational programs, and the implementation of child protection tools in the online environment.

The second paper of this series of articles suggests lifting the veil of secrecy of automated office file formats as a container for manually refining some hidden features. The article discusses an algorithm for implementing one such function — redirecting a hidden link to an active Internet resource. As mentioned earlier in the first part of the article, the increasing number of undeclared functions in the file architecture causes concern among the existing special services of states, which is fully justified by the practical lack of unified software and hardware systems for auditing injections of this kind. Studying simple software tools, unfortunately overlooked by security systems on local PCs and networks, will allow you to build a more powerful closed circuit in the workplace. The existing SIEM in a unified format for building rule bases for closed-circuit auditing does not currently contain ready-made algorithms for detecting extraneous files inside files of automated office formats. This factor needs to be corrected using self-written workplace rules individually. This series of articles does not provide practical instructions for harming the digital environment of the corporate circuit of organizations. The considered scenario of a simple injection with redirection of a link to an Internet resource (redirect function) assumes exclusively educational functions and warnings for information security specialists.

This article provides a comprehensive analysis of the artificial intelligence impact on the processes of digital business transformation in Russia. Key government initiatives, the economic effects of the introduction of artificial intelligence into different economic spheres, industry specifics, regional disparities, as well as challenges and country development prospects are considered. The purpose of the present work is to assess the influence degree of artificial intelligence on the digital transformation of business in Russia, to determine the possibility and prospects for introducing artificial intelligence into improve business processes and increase the competitiveness of Russian companies. The study was carried out using an analysis of the structure and dynamics of the artificial intelligence impact on business development in Russia. Analytical tables and diagrams illustrating the dynamics and scale of changes are provided. Special attention is paid to the role of artificial intelligence as a driver of innovation and increasing the competitiveness of Russian companies in the global digital economy. Artificial intelligence has been introduced into the main processes in 75% of large Russian companies, and among small and medium-sized businesses this figure has reached 50%. The leaders in the artificial intelligence introduction are companies in the financial sector (85%), logistics and retail (70%) and industry (60%). That means that artificial intelligence is no longer an experimental technology and it is becoming an integral part of business processes.

The article discusses modern approaches to predicting crop yields in the agricultural regions of southern Russia using artificial intelligence technologies (neural networks). The relevance of this topic is due to the high importance of the southern regions (Krasnodar Territory, Stavropol Territory, Rostov Region, etc.) in Russia’s food security, and the need for prompt and accurate crop forecasting. The purpose of this work is to develop, apply and evaluate models for predicting crop yields in southern Russia using artificial intelligence methods based on various types of neural networks. Methodology and tools of neural network algorithms application (LSTM, CNN, MLP) are considered to predict crop yields based on data from 2020 to 2025, including statistical indicators of crop yields, meteorological data, and vegetation indices (NDVI). The article presents the results of modeling, which demonstrate the advantage of the LSTM model in terms of prediction accuracy compared to other models. The results section includes graphs and tables that illustrate the actual and predicted crop yields, as well as a comparative analysis of the model errors.

This article proposes a theoretical solution to the problem of overcoming variable uncertainty in leading indicator calculations using economic data expected by business communities as an example. The novelty of the proposed approach lies in its ability to fill a gap in the technology for processing primary data on business community opinions, which is essential for maximizing the utilization of information relevant for decision making. The objective of this article is to present the results of solving a model for encapsulating and decapsulating information with uncertain outcomes. The research method was to construct nonlinear paired regression equations for time series of economic, statistical, and sociological information. The conditions of the model with time series of an independent uncertain variable are examined, and verification and quality assessment of the model are discussed. The study was conducted from 1993 to 2025 using the Bank of Russia and the National Research Financial Institute (NIFI) databases. The data sources included the Moscow Interbank Currency Exchange, Investing, and the analytical departments of commercial banks and brokerage firms. The model was built on a continuous sample of forecast data and opinions from participants in the derivatives markets. The conclusion presents the key results of the model solution, which include a significant increase (by 40%) in the classification accuracy testing for machine learning of the neural network for searching and preprocessing exchange trading data. The advantages of solving a multiple paired regression equation model using a time series of economic indicator values expected by business communities, including a de-encapsulated uncertain variable, are discussed relative to standard solutions of paired regression equations.