micro-F1

A multiclass evaluation metric that aggregates true positives, false positives, and false negatives across all classes to compute global precision and recall, then takes their harmonic mean. It tends to be dominated by classes with large numbers of samples.