tl;dr it's a framework for learning the density of a search space via machine learning and using that to optimize binary search
If you have to run binary search where each probe is expensive, Bayesian Binary Search may be able to help.
Update (July 22, 2025): The paper was published in Algorithms.
Here's an animation I made illustrating the optimization of the search on a normal distribution :). Built with manim.