Tomasz Odrzygóźdź

Dr. Tomasz Odrzygóźdź completed his Ph.D. in Mathematics under the auspices of the Polish Academy of Sciences in 2019. Following this achievement, he joined the artificial intelligence research team led by Dr. habil. Piotr Miłos and Dr. Łukasz Kuciński (AWARElab). During his master’s and doctoral studies, he focused on mathematics and theoretical physics.

Currently, his research revolves around reasoning algorithms in artificial intelligence and reinforcement learning. He is the primary author of the publication titled “Subgoal Search for complex reasoning tasks,” which was featured at NeurIPS 2021. In this work, he applied the mechanism of identifying intermediate goals to effectively solve complex logical problems. He has also authored the paper “Thor: Wielding Hammers to Integrate Language Models and Automated Theorem Provers,” presented at NeurIPS 2022, showcasing the application of language models in conducting mathematical proofs, as well as “Fast and Precise: Adjusting Planning Horizon with Adaptive Subgoal Search,” introducing enhanced versions of search algorithms using intermediate goals.

Tomasz also has a keen interest in physics, particularly in the theory of relativity, quantum mechanics, and cosmology. He has been actively involved in mentoring exceptionally talented youth and has conducted preparatory classes for subject Olympiads. From 2022 to 2023, he was a researcher with the IDEAS NCBR team.