Update README.md

README.md

@@ -88,11 +88,11 @@ The model has been trained on a single epoch spanning about 16000 articles, eval
 
 The model has been evaluated on a random dataset split of 1000 articles not used during training and validation.
 
-- Rouge-1: 33.8%
-- Rouge-2: 17.7%
-- Rouge-L: 31.4%
-- Rouge-Lsum: 31.4%
-- Average length of the generated titles: 12 tokens (about 9 English words)
+- Rouge-1: 37.9%
+- Rouge-2: 24.4%
+- Rouge-L: 35.9%
+- Rouge-Lsum: 35.9%
+- Average length of the generated titles: 13 tokens (about 9 English words)
 
 ### Framework versions