We prove a non-asymptotic generalization of the refined continuity correction developed in Cressie (1978) for the Binomial distribution, which we then use to improve the versions of Tusn\'ady's inequality from Massart (2002) and Carter & Pollard (2004) in the bulk.

翻译：我们证明,在Cressie(1978年)中为Binomial分发开发的精细的连续性校正是非象征性的,我们随后利用这一校正来改进Tusn\'ady在Massart(2002年)和Carter & Pollard(2004年)的不平等。