Machine Learning / AI

In the case of multi-class classification, what are the methods of averaging metrics?

How did you evaluate a multi-agent system? What metrics were used?