Łukasz Kuciński

Hi, I'm Łukasz! My goal is to design artificial intelligence that learns how to solve problems autonomously and in a self-improvement loop.

I am an Assistant Professor at the Polish Academy of Sciences, an Assistant Professor at the University of Warsaw, a Senior Research Scientist at IDEAS, and a member of the ELLIS Society. I am also involved with Ingenix.ai. Recently, I was awarded an OPUS grant for research on debates, Large Language Models (LLMs), and Reinforcement Learning (RL).

I teach Reinforcement Learning and Machine Learning at the University of Warsaw. I regularly review papers for NeurIPS, ICML, ICLR, TMLR, and CoLLAs, including an AC role at NeurIPS24. I am also a member of the Polish AI Olympiad Organizing Committee.

I received my PhD in mathematics from IMPAN and my MSc from the University of Warsaw. Prior to my current position, I led a Solvency 2 internal models team at the Polish Financial Supervision Authority.

See IDEAS website for some PhD research topics.

e-mail: lukasz.kucinski at ideas-ncbr dot pl

X • LinkedIn • Scholar • arXiv • ELLIS • IDEAS

Selected publications

Accelerating Goal-Conditioned RL Algorithms and Research

Michał Bortkiewicz, Władek Pałucki, Vivek Myers, Tadeusz Dziarmaga, Tomasz Arczewski, Łukasz Kuciński, Benjamin Eysenbach

ICLR 2025, spotlight (5% acceptance rate)

BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games

Davide Paglieri, Bartłomiej Cupiał, Samuel Coward, Ulyana Piterbarg, Maciej Wolczyk, Akbir Khan, Eduardo Pignatelli, Łukasz Kuciński, Lerrel Pinto, Rob Fergus, Jakob Nicolaus Foerster, Jack Parker-Holder, Tim Rocktäschel

ICLR 2025

Structured Packing in LLM Training Improves Long Context Utilization

Konrad Staniszewski, Szymon Tworkowski, Yu Zhao, Sebastian Jaszczur, Henryk Michalewski, Łukasz Kuciński, Piotr Miłoś

AAAI 2025

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

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

ICML 2024, spotlight (3.5% acceptance rate)

Magnushammer: A Transformer-based Approach to Premise Selection

Maciej Mikuła, Szymon Antoniak, Szymon Tworkowski, Bartosz Piotrowski, Albert Qiaochu Jiang, Jin Peng Zhou, Christian Szegedy, Łukasz Kuciński, Piotr Miłoś, Yuhuai Wu

ICLR 2024, top-50 average score 8.0

Twitter @Yannik's