Jakub Ivan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jakub Ivan has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- František Duchoň
Institute of Robotics and Cybernetics, Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Ilkovicova 3, 812 19 Bratislava, Slovakia
2 shared publications - Martin Dekan
Institute of Robotics and Cybernetics, Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Bratislava, Slovakia
2 shared publications - Andrej Babinec
Slovak University of Technology
1 shared publication - Eduard Zelenay
Faculty of Electrical Engineering and Information Technology, Slovak University of Technology, Bratislava, Slovakia
1 shared publication - Izabela Trepáčová
Institute of Robotics and Cybernetics, Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Ilkovicova 3, 812 19 Bratislava, Slovakia
1 shared publication - Maroš Kocúr
Faculty of Electrical Engineering and Information Technology, Slovak University of Technology, Bratislava, Slovakia
1 shared publication
Latest Publications
- Open Access
- 6 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (15) , 7849
Large Language Models: From Internal Architecture and Distributed Training Optimisation to Adaptation Strategies
This paper presents a unified technical survey of Large Language Models (LLMs), connecting three layers of the modelling pipeline that existing surveys address in isolation: internal architecture, distributed training optimisation, and downstream adaptation. Its organising principle is the dependency between these layers—how a choice at one constrains what remains feasible at the next. The survey examines fundamental mechanisms (tokenisation, scaled dot-product attention, activation functions, and normalisation, including RMSNorm and pre- versus post-normalisation placement), then the engineering of training at scale: data, tensor, and pipeline parallelism, hybrid schemes, mixed-precision training with BF16 and FP8, ZeRO-Offload memory management, activation checkpointing, and compute-optimal scaling laws together with the conditions under which they fail. The adaptation section covers supervised and instruction fine-tuning, a comparison of parameter-efficient methods (LoRA, QLoRA, adapters, prefix and prompt tuning), Reinforcement Learning from Human Feedback with its reward-hacking failure mode, alternatives including DPO, KTO and Constitutional AI, Retrieval-Augmented Generation beyond the basic pipeline, and decoding strategies. Practical configuration guidance is given for 7B, 70B and trillion-parameter regimes. Dedicated treatments of Mixture-of-Experts architectures, long-context modelling, and hardware-aware co-design close the survey, with open challenges classified by origin and severity.
- 6 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (15) , 7849
Open AccessLarge Language Models: From Internal Architecture and Distributed Training Optimisation to Adaptation Strategies
This paper presents a unified technical survey of Large Language Models (LLMs), connecting three layers of the modelling pipeline that existing surveys address in isolation: internal architecture, distributed training optimisation, and downstream adaptation. Its organising principle is the dependency between these layers—how a choice at one constrains what remains feasible at the next. The survey examines fundamental mechanisms (tokenisation, scaled dot-product attention, activation functions, and normalisation, including RMSNorm and pre- versus post-normalisation placement), then the engineering of training at scale: data, tensor, and pipeline parallelism, hybrid schemes, mixed-precision training with BF16 and FP8, ZeRO-Offload memory management, activation checkpointing, and compute-optimal scaling laws together with the conditions under which they fail. The adaptation section covers supervised and instruction fine-tuning, a comparison of parameter-efficient methods (LoRA, QLoRA, adapters, prefix and prompt tuning), Reinforcement Learning from Human Feedback with its reward-hacking failure mode, alternatives including DPO, KTO and Constitutional AI, Retrieval-Augmented Generation beyond the basic pipeline, and decoding strategies. Practical configuration guidance is given for 7B, 70B and trillion-parameter regimes. Dedicated treatments of Mixture-of-Experts architectures, long-context modelling, and hardware-aware co-design close the survey, with open challenges classified by origin and severity.
- Open Access
- 10 July 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (14) , 7757
A Next-Best-View Method for Complex 3D Environment Exploration Using Robotic Arm with Hand-Eye System
The ability to autonomously generate up-to-date 3D models of robotic workcells is critical for advancing smart manufacturing, yet existing Next-Best-View (NBV) methods often rely on paradigms ill-suited for the fixed-base manipulators found in dynamic industrial environments. To address this gap, this paper proposes a novel NBV method for the complete exploration of a 6-DOF robotic arm’s workspace. Our approach integrates collision-based information gain metric, a potential field technique to generate candidate views from exploration frontiers, and a tunable fitness function to balance information gain with motion cost. The method was rigorously tested in three simulated scenarios and validated on a physical industrial robot. Results demonstrate that our approach successfully maps the majority of the workspace in all setups, with a balanced weighting strategy proving most effective for combining exploration speed and path efficiency, a finding confirmed in the real-world experiment. We conclude that our method provides a practical and robust solution for autonomous workspace mapping, offering a flexible, training-free approach that advances the state-of-the-art for on-demand 3D model generation in industrial robotics.
- 10 July 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (14) , 7757
Open AccessA Next-Best-View Method for Complex 3D Environment Exploration Using Robotic Arm with Hand-Eye System
The ability to autonomously generate up-to-date 3D models of robotic workcells is critical for advancing smart manufacturing, yet existing Next-Best-View (NBV) methods often rely on paradigms ill-suited for the fixed-base manipulators found in dynamic industrial environments. To address this gap, this paper proposes a novel NBV method for the complete exploration of a 6-DOF robotic arm’s workspace. Our approach integrates collision-based information gain metric, a potential field technique to generate candidate views from exploration frontiers, and a tunable fitness function to balance information gain with motion cost. The method was rigorously tested in three simulated scenarios and validated on a physical industrial robot. Results demonstrate that our approach successfully maps the majority of the workspace in all setups, with a balanced weighting strategy proving most effective for combining exploration speed and path efficiency, a finding confirmed in the real-world experiment. We conclude that our method provides a practical and robust solution for autonomous workspace mapping, offering a flexible, training-free approach that advances the state-of-the-art for on-demand 3D model generation in industrial robotics.