Update README.md

README.md

@@ -20,7 +20,7 @@ We fine-tuned DistilBERT on triplets consisting of:
 - a post from a different, randomly selected user (the "negative example")
 
 
-To compute the loss, we use [CLS] encoding of the anchor, positive example and negative example from the last layer of the DistilBERT encoder. We optimize for \\(max(||f(a) - f(n)|| - ||\overline{f(A)} - f(p)|| + \alpha,0)\\)
+To compute the loss, we use [CLS] encoding of the anchor, positive example and negative example from the last layer of the DistilBERT encoder. We optimize for \\(max(||f(a) - f(n)|| - ||f(a) - f(p)|| + \alpha,0)\\)
  
  where:
  - \\( f(a)\\) is the [CLS] encoding of the anchor;