The paper tackles a classic problem in probability theory: how closely does a complicated random variable resemble a standard bell curve (normal distribution), and can we measure that closeness precisely? The specific setting involves random variables built from Poisson processes, which are mathematical objects used to model counts of random events like radioactive decay or arrivals at a queue. The challenge is that earlier results giving sharp numerical bounds on this closeness required extra technical assumptions about the underlying random variables, assumptions that were sometimes hard to verify in practice.
The central contribution is a clever construction called a "martingale core." A martingale is a sequence of probability estimates that updates fairly as new information arrives, like a running average that is always unbiased. The authors show that any random variable in a Poisson chaos (a class built by combining Poisson randomness in structured polynomial-like ways) can be approximated by a sequence of simpler, well-behaved random variables that converge nicely and stay within the same mathematical structure. This approximation acts as a bridge, allowing powerful analytical tools that previously required strong extra assumptions to work under just one mild condition: the random variable must have a finite fourth moment, meaning its extreme fluctuations are not too wild.
Using this framework, the authors prove a clean quantitative result: if such a random variable has variance one and a finite fourth moment, the maximum difference between its distribution and a normal distribution is at most 15.6 times the square root of how far the fourth moment is from 3 (the value the fourth moment takes for a perfect normal distribution). This removes the previously needed technical assumptions from an important 2018 result by Dobler and Peccati, making the bound both simpler to apply and more widely applicable. As a bonus, the paper also establishes refined integrability estimates for related mathematical objects called Malliavin derivatives, which are tools for measuring how sensitive a random variable is to small perturbations in the underlying randomness.