Bartłomiej Cupiał

Hello there! I'm Bartek and I am a PhD student at the University of Warsaw and IDEAS NCBR.

I'm a researcher working on combining reinforcement learning with large language models. Currently, I am exploring how we can use LLMs to supercharge RL algorithms. In particular, how to improve exploration in RL with the help of LLMs and how to integrate external knowledge into RL agents.

When I'm not buried in research papers, you might find me battling in out in NetHack (though I'll admit, my AI is catching up to my skills). I'm always up for a good chat, so feel free to reach out!

News

• I am happy to announce that I will be presenting our paper about fine-tuning in RL at this year's MLinPL conference!. Saturday / 9 November 12:30 - 12:55 Hall A (CfC Session 7) CEST! .
  4 November, 2024
• We talked about work on forgetting in RL fine-tuning at UCL Dark seminar. It was a great experience! You can see the recording here.
  15 July, 2024
• NetHack full moon bug, a story of, by far, the weirdest bug I've encountered in my CS career.
  24 May, 2024
• We will be at ICML 2024! Come by our spotlight poster #1410 on Tue 23 Jul 11:30 a.m. CEST - 1 p.m. CEST! .
  5 May, 2024

Research

• Fine-tuning Reinforcement Learning Models is Secretly a Forgetting Mitigation Problem

  Bartłomiej Cupiał*, Maciej Wołczyk*, Mateusz Ostaszewski, Michał Bortkiewicz, Michał Zając, Razvan Pascanu, Łukasz Kuciński, Piotr Miłoś

  ICML 2024 Spotlight

• GAN-based Plugin Model for Video Generation with Applications in Colonoscopy

  Łukasz Struski1, Tomasz Urbańczyk, Krzysztof Bucki, Bartłomiej Cupiał, Aneta Kaczyńska, Przemysław Spurek, Jacek Tabor