Zanesljiv, enoten, robusten in inteligenten generativni sistem
Duration: September 1, 2025 – August 31, 2028
Project title: Trustworthy Unified Robust Intelligent Generative System
Project acronym: TURING
Project coordinator: Erevnitiko Panepistimiako Institouto Systimaton Epikoinonion kai Ypologiston (ICCS)
Project partners:
RUDOLFOVO ZNANSTVENO IN TEHNOLOSKO SREDISCE NOVO MESTO (RDL), BULL SAS (BULL), UNIVERSITY OF NOVI SAD FACULTY OF SCIENCES (UNSPMF), AEGIS IT RESEARCH GMBH (AEGIS), FONDAZIONE BRUNO KESSLER (FBK), SCUOLA SUPERIORE DI STUDI UNIVERSITARI E DI PERFEZIONAMENTO S ANNA (SSSA), ZELUS IKE (ZELUS), FAKULTETA ZA INFORMACIJSKE STUDIJE V NOVEM MESTU (FIS), DIINEKES S.I. MONOPROSOPI IDIOTIKI KEFALAIOUCHIKI ETAIREIA (DIE), IDRYMA TECHNOLOGIAS KAI EREVNAS (FORTH), NEC LABORATORIES EUROPE GMBH (NEC), IOTAM INTERNET OF THINGS APPLICATIONS AND MULTI LAYER DEVELOPMENT LTD (ITML), ML & AI DATA CONSULTANTS LTD (MLAI), ECOLE DE TECHNOLOGIE SUPERIEURE (ETS), Next Generation Computational Mechanics and Engineering PC (NCOMP), NUCLEAR RESEARCH AND CONSULTANCY GROUP (NRG), UNIVERSITY OF STUTTGART (USTUTT), SLOVENSKÁ TECHNICKÁ UNIVERZITA V BRATISLAVE (STUBA), ORGANISATION EUROPEENNE POUR LA RECHERCHE NUCLEAIRE (CERN), METEO-FRANCE (MF).
Type of project: Horizon Europe RIA – Explainable and Robust AI
Total project value: EUR 7,497,790
Project value for Rudolfovo: EUR 240,000
Project manager at Rudolfovo: Prof. Dr. Simon Muhič (simon.muhic@rudolfovo.eu)
Brief description of the project:
The TURING project is a European research and innovation initiative under the Horizon Europe program (call HORIZON-CL4-2024-HUMAN-03-02 – Explainable and Robust AI), aimed at developing explainable, reliable, and sustainable artificial intelligence (AI) solutions. The main objective of the project is to improve the robustness, explainability, and trustworthiness of generative AI models (genAI) using methods that are mathematically and physically grounded.
The project is developing so-called "TURING models" – generative and multimodal foundation models capable of accurately modeling complex physical phenomena. These models will be part of the "TURING framework," an open-source platform that enables users to interact with models, validate them, and create new applications. The project combines research in the fields of machine learning, physics, computer science, and data science.
TURING addresses key challenges of modern artificial intelligence, such as model reliability in real-world conditions, compliance with legislation (GDPR, AI Act), energy efficiency, explainability of decisions, and prevention of bias. The system uses physically enriched constraints, meta-learning methodologies, federated learning, and distributed computing to improve model robustness.
The results will be applied in three high-tech domains: nuclear energy, particle physics, and meteorology. The project will enable the validation of models on real data, improve prediction accuracy, reduce simulation costs, and accelerate innovation.
The social and economic impacts include democratising access to advanced AI tools, increasing the competitiveness of European companies (especially SMEs), supporting ethical and human-centred development of digital technologies, and strengthening the EU's leadership in artificial intelligence. Through open-source solutions, publications, standardisation and collaboration with organisations such as ADRA, ELLIS and CLAIRE, the project will contribute to the development of trustworthy artificial intelligence in Europe.
The project consortium includes leading European research institutions (e.g., CERN, ETH Zurich, FORTH) and industrial partners (e.g., NEC, Bull, AEGIS). Over a period of three years (36 months), TURING will establish an open, user-friendly platform and develop new methods that will enable the explainable, safe, and effective use of generative artificial intelligence in real-world environments.
