?
JavaCapsule: Итеративная генерация и отладка Java-кода на основе структурированной обратной связи
Труды Института системного программирования РАН. 2026. Т. 38. № 5. С. 73–90.
Alexandrov D., Vasilevsky V., Rezunik L., Kuligin L., Manushkina A., Думкин Н. А., Прозорский М. А., Пинигин К. Ю.
Large language models demonstrate limitations in tasks that require precise information retrieval from very long input sequences. This limitation affects code generation in large object-oriented systems, where correctness depends on complex inter-class dependencies and the analysis of long execution traces that exceed standard context window sizes. This paper presents JavaCapsule, a technology for automated generation and iterative debugging of Java code based on structured feedback. The approach incorporates unit test results and execution traces into an iterative refinement loop. For reliable evaluation, JavaBench+ is introduced as an improved benchmark with corrected structural inconsistencies and extended feedback that includes access to test specifications and execution logs. In addition, the Associative Recurrent Memory Transformer (ARMT) is investigated as a mechanism for accurate information retrieval from long sequences. The study is based on controlled experiments using JavaBench+ and comparative evaluation with a strong baseline model. The results show that JavaCapsule improves Pass@1 by 20 percentage points compared to Google Gemma-3-27B-it on medium-complexity tasks and demonstrates consistent improvements on larger projects. In addition, experiments with ARMT demonstrate accurate information retrieval from sequences of up to one million tokens. The obtained results characterize the capabilities of iterative debugging and long-context processing as two complementary components of the proposed approach.
Keywords: ассоциативная памятьcode generationJavaгенерация кодаязык программирования Javaбольшие языковые модели large language modelsинтеллектуальная отладкамодели с длинным контекстомтрансформеры с расширенной памятьюрекуррентный трансформер ARMTintelligent debugginglong-context modelsmemory-augmented transformersARMT
Publication based on the results of:
Aleksei Samarin, Nazarenko A., Nazarenko D. et al., Machine Learning and Knowledge Extraction 2026 Vol. 8 No. 10 Article 330
Automated microscopic image analysis is essential for quantitative microbiology, yet manual acquisition and generic recognition models struggle with morphological variability and scene artifacts. We address both ends of this pipeline. First, we present a low-cost, open-source acquisition system, based on an Arduino UNO R3 microcontroller and NEMA 17 stepper motors, that automates focusing, magnification, and ...
Added: October 14, 2026
Mamedov I., Fofanov M., Lukianchenko P. et al., IEEE Access 2026 No. 14 P. 26662–26671
DreamBooth-style personalization, which often relies on one or a few images, can memorize context (background, pose, lighting) rather than learning the subject’s identity, thereby reducing fidelity and limiting prompt-driven edits. P-LoRA (Probability LoRA) is introduced as a training-time modification that biases diffusion timestep sampling and utilizes importance weights, ensuring that the overall objective remains unchanged. ...
Added: October 14, 2026
Springer, 2026.
This book highlights cutting-edge research in the field of network science, offering scientists, researchers, students and practitioners a unique update on the latest advances in theory and a multitude of applications. It presents the peer-reviewed proceedings of the XIV International Conference on Complex Networks and their Applications (COMPLEX NETWORKS 2025). The carefully selected papers cover ...
Added: October 13, 2026
Springer, 2026.
This book offers a comprehensive overview of the latest research, innovations, and practical applications in the field of sustainability, with a focus on construction materials and their use in building and civil engineering structures. As an outcome of the 3rd International Conference on Sustainability: Developments and Innovations (ICSDI 2026), this book presents a selection of ...
Added: October 13, 2026
Dmitriev A., Kornilov V., Batienkov R., Lecture Notes in Civil Engineering 2026 Vol. 886 P. 249–255
Numerous studies have provided compelling evidence that the probability distribution of avalanche sizes follows a power law, and that 1/f noise is present in the stochastic dynamics of landslides. We present a theoretical explanation of these phenomena within the framework of self-organized criticality theory, proposing a discrete-time model in the form of a sandpile cellular ...
Added: October 13, 2026
Dmitriev A., Shulman I., Lebedev A., Studies in Computational Intelligence 2026 Vol. 1265 P. 266–273
A discrete-time model is presented for the self-organization of the X/Twitter social network to the edge of a phase transition, characterized by avalanche-like propagation of information in the network. The model is based on the analogy between the critical dynamics of the X social network and sandpile cellular automata with Manna rule on the Chung-Lu ...
Added: October 13, 2026
Makarov D. M., Kalikin N., Fashchevsky K. A. et al., Electrochimica Acta 2026 Vol. 579 Article 150159
Machine learning (ML) has become a powerful tool for accelerating the development of carbon-based supercapacitors;
however, most published models assume that a single predictive framework can describe both conventional
porous carbons and hybrid carbon electrodes. Here, we test this assumption using a comprehensive
benchmark dataset and an extended literature-curated dataset enriched with synthesis- and electrolyte-related
descriptors. Three ML algorithms ...
Added: October 13, 2026
Zhang Z., Ilvovsky D., Information Processing and Management 2026 Vol. 63 No. 7
Transformer-based Large Language Models (LLMs) rely on Key-Value (KV) cache to accelerate autoregressive decoding, but memory and compute requirements still grow with context length, limiting deployment on resource-constrained devices. Existing training-free KV cache compression methods mainly rely on attention-driven retention heuristics, which may under preserve structurally informative tokens under tight cache budgets, especially when such ...
Added: October 13, 2026
No Species Is an Island, Entire of Itself: Genetic Speciation Algorithm with Interplay among Species
Gromov V., Sohrabi M., Fathollahi-Fard A., International Journal of Engineering, Transactions B: Applications 2027 Vol. 40 No. 5 P. 1437–1460
This paper, with a particular emphasis on Genetic Algorithm (GA), proposes a novel approach that integrates clustering mechanisms, predator-prey dynamics, and a distinctive selection strategy for generating subsequent populations. The proposed evolutionary algorithm, referred to as Genetic Speciation Algorithm with Interplay Among Species (GSAIS), is comprehensively evaluated against several variants of GAs as well as ...
Added: October 12, 2026
Singapore: Springer, 2026.
This volume presents the Proceedings of the 34th IUPAP Conference on Computational Physics (CCP2023), held in Kobe, Japan on August 4–8, 2023. It offers a collection of peer-reviewed articles that highlight recent developments in computational and simulation research across various areas of physics. Key topics include the use of data science, quantum computing, and machine ...
Added: October 12, 2026
Zykov S. V., Ермаков С. Р., Современные наукоемкие технологии 2026 № 9 С. 43–53
При федеративном обучении с неоднородными клиентскими данными и переменными задержками запоздавшие обновления могут ухудшать качество модели, а обмен параметрами с глобальным сервером увеличивает сетевую нагрузку. Целью исследования является разработка и теоретико-экспериментальное обоснование метода и алгоритмов адаптивного федеративного обучения с учетом распределенного запаздывания клиентских обновлений. Для обоснования разрабатываемых решений использованы событийная и непрерывная модели градиентных обновлений, ...
Added: October 11, 2026
Брычков М. Е., Незнанов А.А., Программирование 2026 № 4 С. 50–66
The matrix profile (MP) quickly became one of the most important time series preprocessing methods when it was introduced in 2016, facilitating the solution of a wide range of time series analysis problems, in particular anomaly and pattern detection problems. The high significance has led to the emergence of various tools for MP calculating, but ...
Added: October 9, 2026
Aleskerov F. T., Вайншток А. П., Делахова А. М. et al., Информационные процессы 2026 Т. 26 № 3 С. 895–913
The development and effective management of territorial entities are priority issues for all states, particularly in the context of the digital transformation of public administration. Over the past decade, the integration of information technologies into municipal and regional planning has significantly altered approaches to strategic development, service delivery, and public engagement. This paper describes the ...
Added: October 9, 2026
Stepanyants V., Andrey V. Fizulin, Chibirov A. et al., FUTURE TRANSPORTATION 2026 Vol. 6 No. 5 Article 225
Reliable Vehicle-to-Everything (V2X) evaluation requires propagation models that represent terrain and multilevel infrastructure. Most integrated vehicular simulators still rely on planar models, but the magnitude of the resulting bias is unclear. This study quantitatively compares flattened two-dimensional (2D) and terrain-aware three-dimensional (3D) variants of three scenarios using identical Sionna RT settings. The pipeline combines OpenStreetMap ...
Added: October 8, 2026
Люксембург А. А., УРСС, 2005.
Изучается возможность автоматизированного построения математических теорий. Рассматривается дедуктивная система, основанная на языке логики предикатов первого порядка, объектами системы являются математические выражения или формулы, которые описывают математические объекты или их свойства. В дедуктивной системе выводятся математические определения и теоремы. Для доказательства теорем используются методы автоматического доказательства. Разработан алгоритм, выводящий часть формул системы. Для решения задачи используется аппарат математической ...
Added: October 7, 2026
Groshev Maksim, Rybakov Petr, Teterin N. et al., Sensors 2026 Article 6171
Manual assessment of soybean yield components, such as pod number, is laborious, time-consuming, and subjective. Existing computer-vision approaches based on object detection or instance segmentation perform poorly on close-range RGB imagery of soybean canopies due to severe occlusions, ambiguous plant boundaries, and the high cost of bounding-box annotation. To address these challenges, we propose a ...
Added: October 7, 2026
Aleskerov F. T., Chaika E., Дерендяев А. Б. et al., Procedia Computer Science 2026 No. 287 P. 590–595
Under conditions of dynamic socio-economic changes, effective planning and decision-making are key factors in ensuring a high quality of life for the population in territories. This paper presents a decision support system for territorial administrations for the development and implementation of sustainable development strategies, using one of the regions of the Russian Federation – the ...
Added: October 7, 2026
Томск: Издательство Томского государственного университета, 2026.
Материалы сборника Шестнадцатой Международной конференции «Новые информационные
технологии в исследовании сложных структур» (Москва, 21–25 сентября 2026 г.) ориентированы на
широкий круг специалистов, работающих на стыке теории информации, системного анализа и
прикладных предметных областей. В издание вошли результаты исследований, посвящённые
моделированию дискретных и стохастических структур управления и связи, разработке
высокопроизводительных вычислительных и телекоммуникационных систем, а также вопросам
цифровой трансформации образования, архитектурно-градостроительного проектирования,
экологического ...
Added: October 6, 2026
Серебренников Д. А., Belov A. V., Информационные технологии и вычислительные системы 2026 № 3 С. 157–169
В условиях роста стоимости энергоресурсов и необходимости повышения энергоэффективности производственных процессов особую актуальность приобретает задача оптимизации энергопотребления промышленных предприятий. В данной работе рассматривается подход к управлению энергозатратами машиностроительного предприятия на основе методов многокритериальной оптимизации. Постановка задачи включает несколько целевых функций: минимизацию энергопотребления, минимизацию стоимости электроэнергии с учётом тарифных ограничений и максимизацию производственной эффективности. Для решения ...
Added: October 5, 2026
Kuninets A., IEEE Transactions on Information Theory 2026 P. 1–1
In this work we study the applicability of Quasi-Cyclic Subfield Subcodes of Dual Elliptic (QC-SSDE) codes for integration into code-based cryptographic schemes. Detailed algorithms are provided for constructing parity-check matrices as well as block-circulant parity-check matrices for this family of codes, accompanied by empirical results that enable the construction of QC-SSDE codes with predetermined dimensions. ...
Added: October 3, 2026
Sherstinova T., Урих А. Е., Вестник Новосибирского государственного университета. Серия: Лингвистика и межкультурная коммуникация 2026 Т. 24 № 2 С. 42–54
Цифровые технологии XXI века способствуют формированию новых методологических подходов в гуманитарных науках, включая литературоведение и биографические исследования. Для автоматизации процессов сбора и аннотирования информации все чаще используют большие языковые модели (Large Language Models, LLM), включая ChatGPT. Развитие подобных инструментов стимулирует переход от традиционных методологических подходов, основанных на исключительно ручном аннотировании и экспертной верификации, к гибридным, ...
Added: September 28, 2026
Alexandrov D., Vasilevsky V., Rezunik L. et al., В кн.: Труды Института системного программирования РАНТ. 38. Вып. 5.: ИСП РАН, 2026. С. 73–90.
Ограничения при выполнении задач, требующих точного извлечения информации из сверхдлинных последовательностей, приводят к снижению качества генерации и отладки кода крупных объектно-ориентированных систем, характеризующихся сложными межклассовыми зависимостями и необходимостью анализа длинных трассировок выполнения, превышающих размеры стандартных контекстных окон. Предлагается технология JavaCapsule для генерации и отладки Java-кода, реализующая итеративную отладку на основе структурированной обратной связи, включающей результаты ...
Added: September 24, 2026
Тыров И. А., Васильев Ю. А., Владзимирский А. В. et al., Труды Института системного анализа Российской академии наук 2026 Т. 76 № 3 С. 79–89
Large language models (LLMs) in healthcare currently attract significant interest. This study compared GigaChat and YandexGPT by evaluating 1,200 LLM-generated responses to physician queries from outpatient and inpatient settings across 7 topics. Thirteen expert physicians assessed the responses using a 5-point scale across five criteria: relevance, correctness, safety, completeness, and language literacy. We performed paired ...
Added: September 22, 2026
Shipovalova L., Asya A. Filatova, Социология власти 2026 Vol. 38 No. 1 P. 224–243
This article examines the integration of large language model based chatbots into educational practices in Russian higher education through the lens of critical theory and curriculum ideologies. Drawing on Ivan Illich’s concept of convivial tools and Jürgen Habermas’s distinction between instrumental and communicative rationality, the authors conceptualize large language models as technologies capable of operating ...
Added: September 18, 2026