Nedko Savov

Doctoral Student
Email: nedko[dot]savov[at]insait[dot]ai

Website | Scholar | LinkedIn


About me

I am Nedko Savov, a PhD student at the Institute for Computer Science, Artificial Intelligence and Technology, Sofia University. I am supervised by Prof. Luc Van Gool.

I am broadly interested in teaching computers to understand and generate visual worlds – from videos to interactive environments.

My research focuses on advancing interactive world models – generative systems that learn to simulate and create interactive visual worlds. My work explores vision-based generative modeling with an emphasis on interaction, long context processing, memory abilities, and generalization across many environments. 

I am also interested in self-supervised learning, latent action representations, controllable video and image editing, and efficient video generation. Through my past work, I also have a background in face analysis and synthesis, 3D scenes, computational photography.

Academic Service

  • (Neurips’25 Workshop) Embodied World Models for Decision Making – Co-organizer
  • (IOAI’24) International Olympiad of Artificial Intelligence 2024 – Scientific Committee Member

Publications

  • Savov, N., Kazemi, N., Zhang, D., Paudel, D. P., Wang, X., & Van Gool, L. (2025). StateSpaceDiffuser: Bringing Long Context to Diffusion World Models. Advances in Neural Information Processing Systems (NeurIPS’25).
  • Savov, N., Kazemi, N., Mahdi, M., Paudel, D. P., Wang, X., & Van Gool, L. (2025). Exploration-Driven Generative Interactive Environments. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR’25).
  • Akbiyik, M. E., Savov, N., Paudel, D. P., Popovic, N., Vater, C., Hilliges, O., Van Gool, L. & Wang, X. (2025). Leveraging Driver Field-of-View for Multimodal Ego-Trajectory Prediction. The Thirteenth International Conference on Learning Representations (ICLR’25).
  • Kazemi, N., Savov, N., Paudel, D. P., & Van Gool, L. (2024). Learning Generative Interactive Environments by Trained Agent Exploration. NeurIPS 2024 Workshop on Data-driven and Differentiable Simulations, Surrogates, and Solvers (NeurIPS-W’24).
  • Savov, N., Ngô, M., Karaoglu, S., Dibeklioglu, H., & Gevers, T. (2019). Pose and Expression Robust Age Estimation via 3D Face Reconstruction from a Single Image. Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCV-W’19).

Education

  • University of Amsterdam, 09/2016 – 10/2018
    M.Sc. in Artificial Intelligence (Graduated with cum laude)
  • Sofia University, 10/2012 – 06/2016
    B.Sc. in Computer Science (Top of the class)

Relevant Work Experience
Oct 2018 – May 2023 – Scientific Programmer