The Poincare Recurrence Theorem is a classical result in mathematics saying that if you have a system that evolves over time while preserving some notion of volume or probability, almost every point in that system will eventually return arbitrarily close to where it started. The paper asks a more refined question: which specific points are guaranteed to keep returning, not just for one particular transformation, but for all "computable" transformations? A computable transformation is one that can be described by an algorithm, which is a natural constraint when thinking about physically or computationally realizable systems.
To answer this question, the authors develop precise characterizations of which points behave well under recurrence. Some of these characterizations connect to existing notions from algorithmic randomness, a field that formalizes what it means for a number or sequence to be "random" from a computational perspective. Other characterizations require the authors to introduce entirely new concepts they call genericity conditions, which capture a different kind of typicality: roughly, a point is generic in their sense if it cannot be singled out or avoided by sufficiently simple algorithmic procedures. They prove both necessary and sufficient conditions, meaning they identify exactly the right property a point needs to guarantee recurrence across all computable transformations.
The paper also goes beyond just asking whether a point returns to a given region, and investigates how often it returns. The authors examine conditions under which a point revisits regions containing it at a positive long-run frequency, meaning the returns are not just eventual but are in some sense regular. They also study multiple recurrence, a stronger property where the point and several of its iterates all land in the same region simultaneously, which connects to deeper results in ergodic theory and combinatorics. Together, the results give a detailed computational portrait of recurrence behavior in measure-preserving systems.