Victor Prokhorov


My research has revolved around the development of the deep latent variable models with interpretable and structured representations for modelling text and images. The research in this direction can facilitate controlled generation, faithful attribute binding (in the multi-modal setting) and make an analysis of the generative models easier.

Previously, I was a Research Associate (Postdoc) at the University of Edinburgh. Working with Siddharth N. and Ivan Titov on Multimodal Interpretability from Partial Sight.

I also interned at MILA, with Siva Reddy.

I obtained my PhD in Computation, Cognition and Language at the University of Cambridge, Language Technology Lab. I was advised by Ehsan Shareghi and Nigel Collier.

Links to my CV and GitHub.