Tomáš Bureš
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tomáš Bureš has a total of 158 co-authors. Here are the top co-authors by number of shared publications:
- Petr Hnetynka
Charles University
59 shared publications - Ilias Gerostathopoulos
Vrije Universiteit Amsterdam
21 shared publications Frantisek Plasil
Charles University in Prague, Faculty of Mathematics and Physics, Prague, Czech Republic
20 shared publicationsJaroslav Keznikl
Insitute of Computer Science, Academy of Sciences of the Czech Republic, Prague, Czech Republic
17 shared publicationsMichal Kit
Charles University in Prague, Faculty of Mathematics and Physics, Prague, Czech Republic
11 shared publicationsMichal Malohlava
11 shared publications
Latest Publications
- Open Access
- 16 October 2026
- Published by Springer Nature
- p. 107-125
Situation Awareness in AI-Generated Industrial Cyber-Physical Systems
Modern industrial environments increasingly rely on close cooperation between humans and autonomous devices, such as collaborative robots, a relationship that requires reliable situation awareness. At the same time, the AI coding agents (with underlying LLMs) enable a new development paradigm, allowing controllers to be generated from high-level descriptions rather than being implemented manually. While this approach accelerates development, it raises concerns about correctness and safety. In this paper, we propose a method that complements AI-generated controllers with explicit situational descriptions of safety constraints. Instead of addressing safety throughout the control logic, we do so through declarative descriptions of relevant situations and the constraints that must hold in these situations. This enables a clear separation between functional behavior and safety requirements, allowing safety concerns to be expressed independently and concisely. We demonstrate the proposed method on a real-life collaborative human–robot scenario from an ongoing international project.
- 16 October 2026
- Published by Springer Nature
- p. 107-125
Open AccessSituation Awareness in AI-Generated Industrial Cyber-Physical Systems
Modern industrial environments increasingly rely on close cooperation between humans and autonomous devices, such as collaborative robots, a relationship that requires reliable situation awareness. At the same time, the AI coding agents (with underlying LLMs) enable a new development paradigm, allowing controllers to be generated from high-level descriptions rather than being implemented manually. While this approach accelerates development, it raises concerns about correctness and safety. In this paper, we propose a method that complements AI-generated controllers with explicit situational descriptions of safety constraints. Instead of addressing safety throughout the control logic, we do so through declarative descriptions of relevant situations and the constraints that must hold in these situations. This enables a clear separation between functional behavior and safety requirements, allowing safety concerns to be expressed independently and concisely. We demonstrate the proposed method on a real-life collaborative human–robot scenario from an ongoing international project.
- 19 September 2026
- Published by Springer Nature
- p. 57-64
YAMAS: Yet Another Machine Learning Automation System
- 19 September 2026
- Published by Springer Nature
- p. 57-64
YAMAS: Yet Another Machine Learning Automation System
- 27 August 2026
- Published by Springer Nature
- p. 270-278
Towards Component-Based Development for the Era of AI Coding Agents
- 27 August 2026
- Published by Springer Nature
- p. 270-278
Towards Component-Based Development for the Era of AI Coding Agents
- Open Access
- 27 June 2026
- Published by Elsevier in Future Generation Computer Systems
- Vol. 185
On limits of LLMs in adaptation of ensemble-based architectures
- 27 June 2026
- Published by Elsevier in Future Generation Computer Systems
- Vol. 185
Open AccessOn limits of LLMs in adaptation of ensemble-based architectures
- Open Access
- 15 June 2026
- Published by Elsevier in SSRN Electronic Journal
Automated Vibe Coding of a Controller for Collective Adaptive Systems
In collective adaptive systems (CAS), the adaptation manager (AM) is responsible for updating the system’s architecture and behavior at runtime. Recent large language models (LLMs) make it appealing to generate AM code from a system specification, but ensuring correctness is challenging when the code is not manually reviewed.This paper studies fully automated LLM-driven code generation with iterative feedback (vibe coding) for AM implementation. We propose a runtime-verification feedback loop based on Functional Constraints Logic (FCL), a novel temporal logic designed to express fine-grained functional constraints over finite adaptation traces. FCL enables precise, structured violation reports that can be translated into actionable feedback for the LLM. The approach is realised in an end-to-end framework combining vibe-coding loop with feedback based on runtime verification done as part of the adaptation loop.We evaluate the approach on two CAS scenarios. The results show that verification-backed vibe coding can produce valid AMs in a small number of iterations when the functional constraints are stated precisely and violations are reported in a diagnostic form. Detailed, diagnostic feedback is key to eliciting focused code fixes from the LLM.
- 15 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAutomated Vibe Coding of a Controller for Collective Adaptive Systems
In collective adaptive systems (CAS), the adaptation manager (AM) is responsible for updating the system’s architecture and behavior at runtime. Recent large language models (LLMs) make it appealing to generate AM code from a system specification, but ensuring correctness is challenging when the code is not manually reviewed.This paper studies fully automated LLM-driven code generation with iterative feedback (vibe coding) for AM implementation. We propose a runtime-verification feedback loop based on Functional Constraints Logic (FCL), a novel temporal logic designed to express fine-grained functional constraints over finite adaptation traces. FCL enables precise, structured violation reports that can be translated into actionable feedback for the LLM. The approach is realised in an end-to-end framework combining vibe-coding loop with feedback based on runtime verification done as part of the adaptation loop.We evaluate the approach on two CAS scenarios. The results show that verification-backed vibe coding can produce valid AMs in a small number of iterations when the functional constraints are stated precisely and violations are reported in a diagnostic form. Detailed, diagnostic feedback is key to eliciting focused code fixes from the LLM.
- Open Access
- 1 December 2025
- Published by Springer Nature in International Journal on Software Tools for Technology Transfer
- Vol. 27 (6) , 641-655
Understanding ensemble-based component architectures by LLMs
Ensemble-based component systems have been used for many years to develop collective adaptive systems (CAS). The DEECo component model offers a framework for modeling and implementing ensemble-based component systems. Being expressive enough and having semantics specifically tailored towards dynamically evolving systems, DEECo has proven to be fairly powerful in modeling complex and dynamic architectures. We see great potential in employing large language models (LLMs) to simplify creating and refining the DEECo architectures. Since this constitutes a large research scope, in this paper, we focus on initial experiments to demonstrate how well generic LLMs (two OpenAI models executed remotely and four open-source models executed locally) understand the advanced concepts of ensemble-based CAS embodied in DEECo. We do so by systematically asking six questions about specific details of three DEECo applications that differ in the way they are specified. Our results indicate that LLMs can indeed understand ensemble-based architectures and show how this is influenced by the specification means. In particular, using external DSL, which is very self-explanatory, gave good results out of the box. Specifications embedded in existing programming languages needed a prior explanation of how to interpret them.
- 1 December 2025
- Published by Springer Nature in International Journal on Software Tools for Technology Transfer
- Vol. 27 (6) , 641-655
Open AccessUnderstanding ensemble-based component architectures by LLMs
Ensemble-based component systems have been used for many years to develop collective adaptive systems (CAS). The DEECo component model offers a framework for modeling and implementing ensemble-based component systems. Being expressive enough and having semantics specifically tailored towards dynamically evolving systems, DEECo has proven to be fairly powerful in modeling complex and dynamic architectures. We see great potential in employing large language models (LLMs) to simplify creating and refining the DEECo architectures. Since this constitutes a large research scope, in this paper, we focus on initial experiments to demonstrate how well generic LLMs (two OpenAI models executed remotely and four open-source models executed locally) understand the advanced concepts of ensemble-based CAS embodied in DEECo. We do so by systematically asking six questions about specific details of three DEECo applications that differ in the way they are specified. Our results indicate that LLMs can indeed understand ensemble-based architectures and show how this is influenced by the specification means. In particular, using external DSL, which is very self-explanatory, gave good results out of the box. Specifications embedded in existing programming languages needed a prior explanation of how to interpret them.
- Open Access
- 26 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 75-84
A Combined Approach to Performance Regression Testing Resource Usage Reduction
- 26 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 75-84
Open AccessA Combined Approach to Performance Regression Testing Resource Usage Reduction
- Open Access
Towards Continuous Experiment-driven MLOps
Despite advancements in MLOps and AutoML, ML development still remains challenging for data scientists. First, there is poor support for and limited control over optimizing and evolving ML models. Second, there is lack of efficient mechanisms for continuous evolution of ML models which would leverage the knowledge gained in previous optimizations of the same or different models. We propose an experiment-driven MLOps approach which tackles these problems. Our approach relies on the concept of an experiment, which embodies a fully controllable optimization process. It introduces full traceability and repeatability to the optimization process, allows humans to be in full control of it, and enables continuous improvement of the ML system. Importantly, it also establishes knowledge, which is carried over and built across a series of experiments and allows for improving the efficiency of experimentation over time. We demonstrate our approach through its realization and application in the ExtremeXP1 project (Horizon Europe).
Open AccessTowards Continuous Experiment-driven MLOps
Despite advancements in MLOps and AutoML, ML development still remains challenging for data scientists. First, there is poor support for and limited control over optimizing and evolving ML models. Second, there is lack of efficient mechanisms for continuous evolution of ML models which would leverage the knowledge gained in previous optimizations of the same or different models. We propose an experiment-driven MLOps approach which tackles these problems. Our approach relies on the concept of an experiment, which embodies a fully controllable optimization process. It introduces full traceability and repeatability to the optimization process, allows humans to be in full control of it, and enables continuous improvement of the ML system. Importantly, it also establishes knowledge, which is carried over and built across a series of experiments and allows for improving the efficiency of experimentation over time. We demonstrate our approach through its realization and application in the ExtremeXP1 project (Horizon Europe).
- 27 October 2024
- Published by Springer Nature
- p. 208-223
How Well Do LLMs Understand DEECo Ensemble-Based Component Architectures
- 27 October 2024
- Published by Springer Nature
- p. 208-223
How Well Do LLMs Understand DEECo Ensemble-Based Component Architectures
- 7 November 2023
- Published by Springer Nature in International Journal on Software Tools for Technology Transfer
- Vol. 25 (5-6) , 733-746
Generating adaptation rule-specific neural networks
- 7 November 2023
- Published by Springer Nature in International Journal on Software Tools for Technology Transfer
- Vol. 25 (5-6) , 733-746
Generating adaptation rule-specific neural networks