NASK Researchers Presented PLLuM at ML in PL 2026
Researchers from NASK National Research Institute participated in the 10th anniversary edition of ML in PL 2026, one of Poland’s leading conferences dedicated to machine learning and artificial intelligence. The event brought together researchers, industry experts, and practitioners interested in the latest advances and real-world applications of AI technologies.
During the event, a team of researchers from NASK National Research Institute delivered the tutorial “Evaluating LLMs on HPC Clusters: A Step-by-Step Introduction to Slurm with PLLuM”, focused on the evaluation of large language models in high-performance computing (HPC) environments. The hands-on session gave participants practical experience in conducting language model benchmarks using supercomputing infrastructure.
The workshop began with an introduction to PLLuM, a family of Polish large language models developed at NASK. Participants were then introduced to the architecture of HPC systems and the principles of Slurm, the workload manager used on most academic and industrial supercomputers. During the practical part of the tutorial, attendees independently prepared evaluation environments, submitted computational jobs, and analyzed benchmark results conducted on the supercomputing infrastructure of ACK Cyfronet AGH.
The objective of the workshop was to provide participants with the knowledge and practical skills needed to independently perform reliable and reproducible evaluations of language models. Participants gained hands-on experience in managing computational resources, utilizing GPU infrastructure, and analyzing AI model evaluation results. They also generated synthetic data for model post-training and subsequently evaluated the impact of this process on model performance. In addition, participants learned techniques for ensuring experiment reproducibility through proper environment management, configuration tracking, and software versioning.
The tutorial was led by NASK researchers and experts involved in the development of the PLLuM project: Dr. Wojciech Kusa, Head of the Natural Language Processing Department at NASK, Katarzyna Lorenc, Grzegorz Statkiewicz, Marcin Bagnowski, and Dominika Grajewska. Their presentation provided an opportunity to showcase the outcomes of ongoing research at NASK and to exchange experience with the Polish research community working in artificial intelligence and natural language processing.
The workshop was organized as part of the HIVE2 and GAIA AI projects. Computing time on the Athena supercomputer was provided by ACK Cyfronet AGH through the GAIA AI. project.
Participation in ML in PL 2026 highlights NASK’s active role in advancing modern language technologies and strengthening expertise in artificial intelligence, including the development of open Polish large language models.