Congratulations to Brandon Knutson, whose first-author paper On logical extrapolation for Mazes with Recurrent and Implicit Networks was accepted for oral presentation at AAAI-26! His work addresses the question “can certain kinds of neural networks really learn an algorithm?” in the context of a simple problem: find a path through a maze. His thorough and surprising results drew much interest from the reviewers.
AAAI is a top conference in AI, and this year received a record-breaking number of submissions. Approximately 3% were selected for oral presentation. Brandon will travel to Singapore in January to present his work.