Undergrad student, New York University
1 paper at NeurIPS 2025
We propose a one-line modification of the ubiquitous beam search method for graph-based near neighbor search. Our new method yields strong theoretical approximation guarantees and better practical performance.