Abstract. Multidimensional dynamic logic programs (MDLPs) are suitable to represent knowledge dynamic in time, or more generally, information coming from various sources, partially ordered by arbitrary relevancy relation, e.g., level of authority. They have been shown useful for modeling and reasoning about multi-agent systems. Various approaches to define semantics of MDLPs have been presented. Most of the approaches can be characterized as based on rejection of rules. It is understood that on some restricted classes of MDLPs several of these semantics coincide. We focus on acyclic programs. We show that for a MDLP P and a candidate model M, if P is acyclic to some extent then several of the known semantics coincide on M. It follows as a direct consequence that on the class of acyclic programs all of these semantics coincide.