I’m a PhD student at the Department of Informatics, University of Sussex working on deep reinforcement learning and generative models with Chris Buckley and Anil Seth. I’m focusing on probabilistic approaches to control, such as active inference and control-as-inference, and controllable generative modelling. I spent time at Naver Labs Europe working on energy-based models for controllable language generation and am currently a visiting researcher at NYU working with Ethan Perez and Sam Bowman on aligning language models with human preferences.

Before that, I studied cognitive science, philosophy and physics at the University of Warsaw. While there, I worked on compositionality and communication in multi-agent systems at Human Interactivity and Language Lab with Joanna Rączaszek-Leonardi and at AwareLab with Piotr Miłoś as well as on formal models of self-organisation with Marcin Miłkowski.

Highlighted papers

• 

  Controlling conditional language models without catastrophic forgetting

  Tomasz Korbak, Hady Elsahar, Germán Kruszewski, Marc Dymetman

  Accepted for ICML 2022

  ArXiv

• 

  Catalytic role of noise and necessity of inductive biases in emergence of compositional communication

  Łukasz Kuciński, Tomasz Korbak, Paweł Kołodziej, Piotr Miłoś

  NeurIPS 2021

  arXiv

• 

  Energy-based models for code generation under compilability constraints

  Tomasz Korbak, Hady Elsahar, Marc Dymetman, Germán Kruszewski

  NLP4Programming workshop, ACL 2021

  arXiv

• 

  Measuring non-trivial compositionality in emergent communication

  Tomasz Korbak, Julian Zubek, Joanna Rączaszek-Leonardi

  Emergent communication workshop, NeurIPS 2020

  arXiv code

• 

  Developmentally motivated emergence of compositional communication via template transfer

  Tomasz Korbak, Julian Zubek, Łukasz Kuciński, Piotr Miłoś, Joanna Rączaszek-Leonardi

  Emergent communication workshop, NeurIPS 2019

  arXiv code