Research
ELLIS fosters international collaboration across domains, connecting top researchers while investing in the next generation of AI talent.
Members
ELLIS Members are leading scientists in machine learning and AI, shaping Europe's global position in these fields.
About
ELLIS is a pan-European AI network of excellence built upon machine learning as the driver for modern AI.
Funded ELLIS PhD: Multimodal Physically-Consistent World Modeling
Location: Inria Paris, France (Astra-Vision group)
Funding: PEPR AI (3 years)
Supervisors: Fabio Pizzati and Raoul de Charette
Starting date: Spring 2027 (approx.)
This PhD aims to develop multimodal world models that better understand and predict physical interactions, exploiting complementary information from vision, audio, and other sensory modalities.
The research will explore three main directions:
- Understanding physical knowledge: Investigate how physical properties are learned, represented, and shared across modalities in existing world models.
- Improving physical consistency: Develop multimodal learning strategies that leverage complementary sensory information to improve physical realism and predictions.
- Scaling multimodal learning: Explore efficient and automated approaches to combining sensory modalities for physically grounded world modeling.
Research
About
