Cognitive Adaptation of Sliding Window Parameters of Large Language Models Based on a High-Speed Area Ratio Method
Main Article Content
Abstract
This paper proposes a method for adaptive control of sliding window parameters in large language models (LLMs) based on Mamdani fuzzy inference with defuzzification using the high-speed area ratio method (MAR-II). The input variables are the type-token ratio (TTR) and the average BPE token length (ATL), interpreted as measurable correlates of the cognitive load of a text fragment. A nine-rule fuzzy controller generates the sliding window stride and context length (max_length). MAR-II is compared with the center of gravity (CoG) method across a 25×25 grid of input space points (RMSE: 99.9 tokens for stride, 147.4 tokens for max_length). It is demonstrated that MAR-II ensures the additivity of the fuzzy model and eliminates the systematic errors of CoG, achieving a single-inference computation time of approximately 35 ns on a Xilinx Spartan 3E FPGA.
Article Details
References
2. Bobyr M.V., Milostnaya N.A., Kulabuhov S.A. A method of defuzzification based on the approach of areas' ratio // Applied Soft Computing. 2017. Vol. 59. P. 19–32. https://doi.org1 0.1016/j.asoc.2017.05.040
3. Bobyr M.V., Milostnaya N.A., Arkhipov A.E., Luneva M.Yu. Analysis of high-speed defuzzifiers in the problem of robot manipulator control // Vestnik Komp'yuternykh i Informatsionnykh Tekhnologiy. 2020. Vol. 17, No. 4. P. 18–28. https://doi.org1 0.14489/vkit.2020.04.pp.018-028
4. Bobyr M.V., Milostnaya N.A., Belskaya S.Yu. Fuzzy-logic adaptation of sliding window parameters in data preparation for large language models (LLMs) // Elektronnye Biblioteki [Electronic Libraries]. 2026. Vol. 29, No. 4. P. 1–21.
5. Baddeley A. Working memory // Science. 1992. Vol. 255. P. 556–559. https://doi.org1 0.1126/science.1736359
6. Miller G.A. The magical number seven, plus or minus two // Psychological Review. 1956. Vol. 63(2). P. 81–97. https://doi.org1 0.1037/h0043158
7. Bobyr M.V., Yakushev A.S., Dorodnykh A.A. Fuzzy devices for cooling the cutting tool of the CNC machine implemented on FPGA // Measurement. 2020. Vol. 152. P. 107378. https://doi.org1 0.1016/j.measurement.2019.107378
8. Bobyr M.V., Milostnaya N.A., Bulatnikov V.A. The fuzzy filter based on the method of areas' ratio // Applied Soft Computing. 2022. Vol. 117. P. 108449. https://doi.org1 0.1016/j.asoc.2022.108449
9. Bondar O.G., Brezhneva E.O., Golubev D.A. Selection of bit width for nonlinear neuron components in FPGA implementation // Izvestiya Yugo-Zapadnogo Gosudarstvennogo Universiteta [Proceedings of the Southwest State University]. 2025. Vol. 29, No. 4. P. 70–92. https://doi.org1 0.21869/2223-1560-2025-29-4-70-92.
This work is licensed under a Creative Commons Attribution 4.0 International License.
Presenting an article for publication in the Russian Digital Libraries Journal (RDLJ), the authors automatically give consent to grant a limited license to use the materials of the Kazan (Volga) Federal University (KFU) (of course, only if the article is accepted for publication). This means that KFU has the right to publish an article in the next issue of the journal (on the website or in printed form), as well as to reprint this article in the archives of RDLJ CDs or to include in a particular information system or database, produced by KFU.
All copyrighted materials are placed in RDLJ with the consent of the authors. In the event that any of the authors have objected to its publication of materials on this site, the material can be removed, subject to notification to the Editor in writing.
Documents published in RDLJ are protected by copyright and all rights are reserved by the authors. Authors independently monitor compliance with their rights to reproduce or translate their papers published in the journal. If the material is published in RDLJ, reprinted with permission by another publisher or translated into another language, a reference to the original publication.
By submitting an article for publication in RDLJ, authors should take into account that the publication on the Internet, on the one hand, provide unique opportunities for access to their content, but on the other hand, are a new form of information exchange in the global information society where authors and publishers is not always provided with protection against unauthorized copying or other use of materials protected by copyright.
RDLJ is copyrighted. When using materials from the log must indicate the URL: index.phtml page = elbib / rus / journal?. Any change, addition or editing of the author's text are not allowed. Copying individual fragments of articles from the journal is allowed for distribute, remix, adapt, and build upon article, even commercially, as long as they credit that article for the original creation.
Request for the right to reproduce or use any of the materials published in RDLJ should be addressed to the Editor-in-Chief A.M. Elizarov at the following address: amelizarov@gmail.com.
The publishers of RDLJ is not responsible for the view, set out in the published opinion articles.
We suggest the authors of articles downloaded from this page, sign it and send it to the journal publisher's address by e-mail scan copyright agreements on the transfer of non-exclusive rights to use the work.