Bartlomiej "Bart" Sobieski
Researcher in generative modeling and explainability.
I am a second-year Ph.D. student in Computer Science at the University of Warsaw, under the supervision of Prof. Przemysław Biecek and co-supervision of Dr. Matthew Tivnan. My research focuses on developing generative modeling techniques to construct novel explainability algorithms for computer vision models. My greatest passion lies in exploring advanced mathematical methodologies and translating them into scalable practical use cases.
Currently, I serve as an AI Researcher and Focus Lead at the Centre for Credible AI. Recently, I completed a Visiting Researcher tenure at Harvard Medical School & Massachusetts General Hospital.
selected publications
- Auditing Sybil: Explaining Deep Lung Cancer Risk Prediction Through Generative Interventional AttributionsIn International Conference on Machine Learning (ICML), 2026
- System-Embedded Diffusion Bridge ModelsIn Advances in Neural Information Processing Systems (NeurIPS), 2025
- Rethinking Visual Counterfactual Explanations Through Region ConstraintIn International Conference on Learning Representations (ICLR), 2025
- Global Counterfactual DirectionsIn European Conference on Computer Vision (ECCV), 2024