All issues

This paper presents an experimental study of the application of automatic annotation of text data in the question – answer format (QA pairs) under conditions of limited computing resources and data protection requirements. Unlike traditional approaches based on rigid rules or the use of external APIs, we propose using small language models with a small number of parameters that can function locally without a GPU on standard CPU systems. Two models were selected for testing — Gemma-3-4b and Qwen-2.5-3b (quantized 4-bit versions) — and a corpus of documents with a clear structure and a formally rigorous style of presentation was used as source material. An automatic annotation system was developed that implements the full cycle of QA dataset generation: automatic division of the source document into logically connected fragments, formation of “question – answer” pairs using the Gemma-3-4b model, preliminary verification of their correctness using Qwen-2.5-3b based on evidence span from the context and expert quality assessment. The results are exported in JSONL format. Performance evaluation covers the entire QA pair generation system, including fragment processing by the local language model, text preprocessing and postprocessing modules. Performance is measured by the time it takes to generate a single QA pair, the total throughput of the system, RAM usage, and CPU load, which allows for an objective assessment of the computational efficiency of the proposed approach when running on a CPU. An experiment on an extended sample of 12 documents showed that automatic annotation demonstrates stable performance when processing different types of documents, while manual annotation is characterized by significantly higher time costs and high variability. Depending on the type of document, the acceleration of annotation compared to the manual process ranges from 8 to 14 times. Quality analysis showed that most of the generated QA pairs have high semantic consistency with the original context, with only a limited proportion of data requiring expert correction or exception. Although full manual validation of the corpus (the “gold standard”) was not performed as part of this work, the combination of automatic evaluation and selective expert review allows us to consider the resulting quality level acceptable for preliminary automated annotation tasks. Overall, the results confirm the practical applicability of small language models for building autonomous and reproducible automatic text annotation systems under limited computational resources and provide a basis for further research in the field of effective training corpus preparation for natural language processing tasks.

We describe an engineering-psychological experiment that continues the study of ways to adapt a person to the increasing complexity of logical problems by presenting a series of problems of increasing complexity, which is determined by the volume of initial data. Tasks require calculations in an associative or non-associative system of operations. By the nature of the change in the time of solving the problem, depending on the number of necessary operations, we can conclude that a purely sequential method of solving problems or connecting additional brain resources to the solution in parallel mode. In a previously published experimental work, a person in the process of solving an associative problem recognized color images with meaningful images. In the new study, a similar problem is solved for abstract monochrome geometric shapes. Analysis of the result showed that for the second case, the probability of the subject switching to a parallel method of processing visual information is significantly reduced. The research method is based on presenting a person with two types of tasks. One type of problem contains associative calculations and allows a parallel solution algorithm. Another type of problem is the control one, which contains problems in which calculations are not associative and parallel algorithms are ineffective. The task of recognizing and searching for a given object is associative. A parallel strategy significantly speeds up the solution with relatively small additional resources. As a control series of problems (to separate parallel work from the acceleration of a sequential algorithm), we use, as in the previous experiment, a non-associative comparison problem in cyclic arithmetic, presented in the visual form of the game “rock, paper, scissors”. In this problem, the parallel algorithm requires a large number of processors with a small efficiency coefficient. Therefore, the transition of a person to a parallel algorithm for solving this problem is almost impossible, and the acceleration of processing input information is possible only by increasing the speed. Comparing the dependence of the solution time on the volume of source data for two types of problems allows us to identify four types of strategies for adapting to the increasing complexity of the problem: uniform sequential, accelerated sequential, parallel computing (where possible), or undefined (for this method) strategy. The Reducing of the number of subjects, who switch to a parallel strategy when encoding input information with formal images, shows the effectiveness of codes that cause subject associations. They increase the speed of human perception and processing of information. The article contains a preliminary mathematical model that explains this phenomenon. It is based on the appearance of a second set of initial data, which occurs in a person as a result of recognizing the depicted objects.

The numerical methods used to simulate the evolution of hydrodynamic systems require the considerable use of computational resources thus limiting the number of possible simulations. The data-driven simulation technique is one promising approach to the development of heuristic models, which may speed up the study of such models. In this approach, machine learning methods are used to tune the weights of an artificial neural network that predicts the state of a physical system at a given point in time based on initial conditions. This article describes an original neural network architecture and a novel multi-stage training procedure which create a heuristic model of a two-phase flow in a heterogeneous porous medium. The neural network-based model predicts the states of the grid cells at an arbitrary timestep (within the known constraints), taking in only the initial conditions: the properties of the heterogeneous permeability of the medium and the location of sources and sinks. The proposed model requires orders of magnitude less processor time in comparison with the classical numerical method, which served as a criterion for evaluating the effectiveness of the trained model. The proposed architecture includes a number of subnets trained in various combinations on several datasets. The techniques of adversarial training and weight transfer are utilized.

The paper considers the application of swarm intelligence technology to build agent-based simulation models. As an example, a minimal model is constructed illustrating the influence of information influences on the rules of behavior of agents in the simplest model of competition between two populations, whose agents perform the simplest task of transferring a resource from a mobile source to their territory. The algorithm for the movement of agents in the model space is implemented on the basis of the classical particle swarm algorithm. Agents have a life cycle, that is, the processes of birth and death are taken into account. The model takes into account information processes that determine the target functions of the behavior of newly appeared agents. These processes (training and poaching) are determined by information influences from populations. Under certain conditions, a third population arises in the agent system. Agents of such a population informatively influence agents of other populations in a certain radius around themselves, changing.

As a result of the conducted simulation experiments, it was shown that the following final states are realized in the system: displacement of a new population by others, coexistence of a new population and other populations and the absence of such a population. It has been shown that with an increase in the radius of influence of agents, the population with changed rules of behavior displaces all others. It is also shown that in the case of a hard-to-access resource, the strategy of luring agents of a competing population is more profitable.

The problem of allocation of Steiner points in Euclidean Steiner Tree is considered. The cost of network is sum of building costs and cost of the information transportation. Euclidean Steiner tree problem in the form of topological network design is a good model of this problem.

The package MatLab has the way to solve the second part of this problem — allocate Steiner points under condition that the adjacency matrix is set. The method to get solution has been worked out. The Steiner tree is formed by means of solving of the sequence of "three points" Steiner