Laurent Dinh

Papers

Total Citations

100

H-Index

About

Laurent Dinh is a leading researcher at the intersection of generative modeling and representation learning, with a focus on developing models that capture the structure of complex, high-dimensional data. His work is particularly known for advancing flow-based generative models—a class of models that learn invertible transformations between data and latent spaces. Dinh’s most cited paper, "VideoFlow: A Flow-Based Generative Model for Video" (2019, 87 citations), introduced a novel approach to modeling video sequences by learning to predict future frames, enabling the model to capture physical interactions and temporal dynamics. This work has been influential in pushing generative models beyond static images into the domain of video prediction. Another notable contribution is "Learning Awareness Models" (2018, 13 citations), where Dinh explored how agents can learn representations of external objects purely from proprioceptive feedback—a step toward embodied AI. His research has been recognized for its originality and technical depth, earning him a reputation as a key figure in the development of scalable, principled generative models. Dinh’s work continues to inspire new directions in unsupervised learning, video generation, and model-based reinforcement learning.