The University of Illinois Urbana-Champaign Department of Statistics congratulates Eren Kizildag on receiving the Best Paper Award at the Algorithmic Learning Theory Conference 2026, held February 23–26 in Toronto, Canada.
Kizildag’s paper, “The Planted Number Partitioning Problem,” explores a fundamental computational problem involving how to divide a set of numbers into two groups with nearly equal sums. The work introduces a new “planted” model of the problem and reveals surprising theoretical limits on how efficiently algorithms can identify optimal solutions.
The results provide the first demonstration that a structural phenomenon known as the multi-Overlap Gap Property can prevent stable algorithms from solving certain planted problems efficiently, offering new insight into the limits of algorithms used in machine learning and statistics.