Computer EngineeringKNU · FRECS · Department

Long-term memory and evaluation of large language models

High-performance computing, digital communications and artificial intelligence.

Research group of the Department of Computer Engineering

Group lead

Pavlo Biletskyi, PhD in Computer Engineering, Assistant Professor at the Department of Computer Engineering.

Members

  • Leonid Chepel — Doctor of Philosophy (PhD); responsible executor of the project
  • thesis students of the educational and professional programme “Computer Systems and Networks Engineering” (see the list of theses below)

Research area

The group studies methods for extending the context window of large language models through external long-term memory: block-based organisation of stored information with multi-tier access, selective memorisation of information that is new relative to already acquired knowledge, and asynchronous memory management by a separate background process. The scientific novelty lies in a persistent multi-tier block memory outside the context window, fine-tuned on top of existing open models with parameter-efficient methods, without changing their weights and without training a model from scratch.

A second line of work concerns honest evaluation: Ukrainian-language benchmarks for LLMs, isolating the contribution of memory from that of the context window, and evaluation protocols without literal lexical overlap between question and text.

The research is carried out on open models of the Gemma and Qwen families using GPU accelerators.

Project

“Development of a scalable long-term memory architecture for large language models to process ultra-long contexts”. Principal investigator: P. V. Biletskyi. Submitted to the young-researcher competition of the Ministry of Education and Science of Ukraine (2026); proposed term 01.01.2027 — 31.12.2028. The application is under review.

Qualification theses completed within this research area

2026

  • Antoniuk I. V. “An intelligent assistant for a word processor based on cloud and local large language models” — project repository
  • Kulik A. K. “Development of a software system that processes non-deterministic elements using LLM agents for automated analysis and evaluation of NPL portfolios”
  • Mikhnenko V. M. “Development of a video stream relay system with monitoring and diagnostic tools”
  • Palchuk A. M. “Development of a comprehensive benchmark for evaluating the quality of large language models’ performance in the Ukrainian language”
  • Solovei O. S. “Development of an intelligent assistant for spreadsheet processors based on open large language models” — project repository
  • Tyshchenko Ye. V. “Development of a training dataset for the internal memory system of large language models”

2025

  • Khimin R. O. “Improving the performance of large language models on tasks in the Ukrainian language”

Type at least 2 characters

↑↓ selectEnter openEsc close/ search from any page