Adam R. Kosiorek
Papers
2
Total Citations
145
H-Index
2
About
Adam R. Kosiorek is a leading researcher in generative modeling and object-centric representation learning, with a focus on enabling machines to perceive and reason about visual scenes in a compositional manner. His most influential work, the GENESIS framework (Generative Scene Inference and Sampling with Object-Centric Latent Representations), has garnered over 145 combined citations across its 2019 and 2020 publications. This work addresses a critical limitation in standard generative models: their inability to explicitly capture the distinct objects that compose natural scenes. By introducing a latent-variable model that learns to segment and represent individual objects without supervision, Kosiorek’s contributions have paved the way for more robust scene understanding in robotics and reinforcement learning. His research sits at the intersection of computer vision, probabilistic machine learning, and embodied AI, offering a principled approach to decomposing complex visual input into interpretable, object-centric components. Kosiorek’s work is widely recognized for advancing the field’s ability to build compositional world models, a key step toward more general and sample-efficient artificial intelligence.
Research Focus
Key Achievements
Top Papers
- 1
- 2