Update README.md

README.md

@@ -90,7 +90,8 @@ This model was evaluated on the following datasets:
 #### Metrics
 
 <!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
+We provide the macro-averaged F1 scores here; given that all of the downstream token classification tasks in these experiments show significant class imbalance, the weighted-average scores tend to be uniformly higher than their macro-averaged counterparts.
+In the interest of more fairly representing the less prevalent classes and highlighting the difficulty of capturing the long-tailed nature of the distributions in these datasets, we stick to the macro average.
 
 ### Results