We are happy to announce that PREMISE (PREdictive Multimodal Intelligence with Spatial and Episodic memory) has been funded by the Italian Ministry of University and Research (MUR) under the PRIN 2026 call, with a total budget of about €1.46 million. The project is coordinated by Lorenzo Baraldi (AImageLab, UNIMORE) and ranked second in the PE6 panel (Computer Science and Informatics).
PREMISE will develop multimodal foundation models that perceive, remember and reason in 3D space and over long time horizons. The project combines four research lines:
- Semantic perception: open-vocabulary understanding of people, objects, actions and events;
- Episodic memory: consolidation, retrieval and semantic abstraction of experience over long horizons;
- Spatial cognition: a persistent 3D scene model in which past and current states are represented together;
- Predictive intelligence: surprise-driven, uncertainty-aware agentic control.
Trustworthiness is a cross-cutting concern throughout the project.
PREMISE is a three-year effort carried out by five research units: the University of Modena and Reggio Emilia (coordinator), the University of Bologna (Matteo Poggi), the University of Trento (Massimiliano Mancini), CNR-ISTI (Fabio Carrara) and Politecnico di Milano (Federica Arrigoni). The consortium includes members of three ELLIS Units (Modena, Trento and Milan) and strengthens the
collaboration among Italian ELLIS sites on multimodal and embodied AI. Several industrial partners support the project, including AMD Silo AI, whose Physical AI leadership has welcomed the project and its research agenda.
Stay tuned for open positions on the project!