Socially assistive robotics (SAR) is a growing area of research. Evaluating SAR systems presents novel challenges. Using a robot for a socially assistive task can have different benefits and ethical implications. Many questions are important to understanding whether a robot is effective for a given application domain. This paper describes several benchmarks for evaluating SAR systems. There exist numerous methods for evaluating the many factors involved in a robot's design. Benchmarks from psychology, anthropology, medicine, and human-robot interaction are proposed as measures of success in evaluating a given SAR system and its impact on the user and broader population.