Research Internship: Foundations of generative modeling

With the support of researchers from Criteo AI Lab and Inria, the recruited person will:

• Explore theoretical connections between different families of generative models
• Implement hybrid prototypes combining the strengths of GANs and diffusion models
• Develop robust evaluation protocols to quantify improvements
• Contribute to the development of novel theoretical frameworks`

For better knowledge of the proposed research topic: Relevant scientific references are available in our recent publication "Unifying GANs and score-based diffusion as generative particle models" (NeurIPS 2023) and "Improving consistency models with generator-augmented flows" (arXiv 2024).

Proposed Research Directions

• Scaling hybrid GAN-diffusion models from prototype to state-of-the-art quality
• Extending theoretical links between diffusion and GANs by removing simplifying hypotheses
• Investigating the gap between one-step generator training and diffusion distillation
• Extending diffusion distillation to more general multi-step models like flow matching

• Strong background in machine/deep learning and mathematics
• Knowledge of computer vision and image processing
• Familiarity with generative models (GANs, diffusion, consistency models) and deep learning architectures (CNNs, transformers)
• Proficiency in Python and experience with deep learning frameworks (PyTorch and/or JAX)
• Appetite for theoretical and applied scientific research
• Ability to conduct exploratory research with significant autonomy

• Subsidized meals
• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities