Mathematicians studying the foundations of set theory are interested in "large cardinal axioms," which assert the existence of extraordinarily large infinite sets with special symmetry properties. Two recently introduced varieties, called exacting and ultraexacting cardinals, sit near the very top of the known large cardinal hierarchy. This paper investigates how these cardinals behave when the universe of sets is modified through a technique called forcing, which is a standard tool for building new mathematical universes from existing ones while carefully controlling which statements become true or false.
Several concrete results are established. First, the authors show that starting from a sufficiently strong large cardinal assumption (called an I2-embedding), one can construct a well-behaved model of set theory containing a proper class of exacting cardinals. They also show that ultraexacting cardinals can consistently accumulate, meaning a single cardinal can sit above a whole sea of ultraexacting cardinals below it, but that exacting cardinals cannot accumulate in this way if a prominent open conjecture due to Woodin, called the HOD Conjecture, is true. This conjecture concerns a canonical inner universe called HOD, which collects all sets that are "ordinal definable," and plays a central role throughout the paper.
The paper also examines how exacting cardinals appear inside HOD itself. If an exacting cardinal exists and a related technical condition holds, that cardinal turns out to carry strong large cardinal properties even within the more restricted HOD universe. Additionally, the authors construct a model where the cardinal directly above an exacting cardinal enjoys remarkable definability properties across many different HOD-like sub-universes simultaneously. Together, these results deepen understanding of how the most extreme large cardinal notions interact with the structure of the set-theoretic universe and with the foundational question of how much of mathematics is ultimately reducible to ordinal definability.