Supporting Public Administration
We develop and implement large language models (LLMs), drawing on the experience gained during the implementation of the PLLuM project.
Scope of work
The Project Includes Four Complementary Activities
Building language data corpora for pretraining, fine-tuning, and alignment of large language models
We are collecting new text datasets, especially publicly unavailable data and documents from the administrative domain. We are also preparing instruction datasets for fine-tuning and preference data for model alignment based on our own typology.Training large language models, including pretraining, fine-tuning, and alignment
We plan to expand the PLLuM family of models with various sizes, including general-purpose models adapted to diverse NLP tasks — with a strong focus on the administrative domain — as well as generative models, including advanced RAG-based models (Retrieval Augmented Generation).Securing and evaluating large language models
We are developing tools for comprehensive quality and security evaluation of language models in various government use cases. To reduce the risk of harmful or undesired content, we are building input/output filtering algorithms and implementing output correction mechanisms.Pilot implementation of models in the public sector
We support the preparatory process for deploying Polish language models in the mObywatel app (including creating a mObywatel chatbot) and in selected public institutions, such as the Ministry of Digital Affairs, in the form of virtual assistant pilots.
About the consortium
Meet the People Behind HIVE
IT specialists, linguists, as well as lawyers, sociologists, and cybersecurity experts. We all work together to advance Polish language models.
Funding
The Project Is Funded by the Ministry of Digital Affairs
The project is implemented under a targeted grant No. 1/WI/DBI/2025, titled HIVE AI: Development and pilot implementation of large language models in the Polish public administration sector. Total funding amount: PLN 18,983,055.
Contact
Feel Free to Get in Touch!
Office
Please send all correspondence to the NASK headquarters in Warsaw.Kolska Street 12, 01-045 Warszawa