Hodge integrals are certain numbers that arise in the mathematical study of the geometry of curves (one-dimensional geometric objects). They come from integrating specific quantities over spaces that parameterize all curves of a given genus (a measure of complexity, roughly the number of holes). These integrals produce rational numbers, meaning fractions, and a natural question is: what denominators can appear? The least common denominator of all such fractions in a given genus tells you something deep about the arithmetic structure underlying the geometry.
The paper proves that this least common denominator is exactly a quantity called the Minkowski number, which can also be described as a generalization of the factorial function called the Bhargava factorial, defined using prime numbers. Minkowski numbers are classical objects in number theory, so connecting them to geometry is a meaningful and surprising result. As a consequence, the authors resolve a conjecture by Liu and Xu about how often the prime number 3 divides these denominators, and they extend a result related to an older conjecture by Itzykson and Zuber. Specifically, the order of the symmetry group of any sufficiently simple curve always divides the least common denominator of Hodge integrals at the relevant genus, which is a precise statement about how geometry and arithmetic constrain each other.
The paper also includes an unusual methodological note: the authors used an AI benchmarking tool called IMProofBench during the research process. They are transparent that the final paper is entirely human-written, but they found the AI tool genuinely useful in the development of the work, and they reflect briefly on what that experience revealed about the current capabilities and limitations of AI assistance in mathematical research.