Community detection, detecting groups of densely connected vertices in a graph, is a fundamental problem in unsupervised network science. Typical community detection algorithms make the strong assumption that each vertex must belong to exactly one community; that is, the communities form a partition. In this talk, we relax this restriction by allowing vertices to be part of many communities (overlap) or no communities (outliers). We cover an extension of the Artificial Benchmark for Community Detection (ABCD), a synthetic graph model with known community structure, to have overlaps and outliers. We will also discuss a recent method for comparing the detected communities to the known communities, which is used to benchmark community detection algorithms.