Rishi Hazra Successfully Defends PhD Thesis on Neurosymbolic Decision-Making with LLMs
By integrating LLMs with various logic-based frameworks, Rishi Hazra demonstrated how these models can be enhanced to reason and plan more effectively.
On 10 October, Rishi Hazra successfully defended his doctoral thesis titled “Neurosymbolic Decision-Making with Large Language Models” at Örebro University. The defense was met with unanimous praise from the jury, who commended the quality and depth of his work, which is based on an impressive portfolio of 13 publications, five of which were included in the compilation thesis.
Opponent: Prof. Scott Sanner, University of Toronto
Committee: Steven Schockaert, professor, Cardiff University, Moa Johansson, docent, Chalmers tekniska högskola, Sven Koenig, professor, University of California, Irvine and WASP guestprofessor at Örebro.
Hazra’s research explored whether large language models (LLMs) are capable of reasoning. This question led him to develop novel neuro-symbolic approaches that significantly outperformed existing state-of-the-art methods. By integrating LLMs with various logic-based frameworks, Hazra demonstrated how these models can be enhanced to reason and plan more effectively.
He is the fourth PhD student funded by the Wallenberg AI, Autonomous Systems and Software Program (WASP) at Örebro University to defend his thesis, following Tiago Almeida, Quantao Yang, and Mohamadreza Faridghasemnia. Hazra is currently employed at Meta.
During the post-defense speeches, the supervisor Luc de Raedt remarked on Hazra’s transformation after his first internship at Meta, noting how the experience deepened his appreciation for formalization. Pedro Zuidberg dos Martirez, Hazra’s second supervisor, humorously reflected on the brilliance of the thesis, saying he often wondered whether Hazra was a genius or just a bit crazy—ultimately leaning toward genius.
Hazra expressed thanks to everyone who supported him throughout his journey, including his mentors at Meta.
Text: Carolina Wittenfeldt
Photo: Franziska Klügl