Update README.md

README.md

@@ -79,10 +79,10 @@ The corpus is based on the dataset introduced by Pang and Lee (2005) and consist
 
 ## Dataset Structure
 
-The modifications carried out in the dataset are thought to evaluate cognitive biases in a few-shot setting and with different task complexities. 
+The modifications carried out in the dataset are thought to evaluate cognitive biases in a few-shot setting and with two different task complexities. 
 We make use of 25,000 instances from the original dataset, while the remaining ones serve as few-shot examples.
 Each instance is prompted with all possible unbalanced 4-shot distributions.
-We also increase task complexity by introducing an additional neutral example between the first and last two examples.
+We increase task complexity by introducing an additional neutral example between the first and last two examples.
 
 **Dataset Fields**
 - `idx`: original sentence id, in the format `<original_partition>_<original_id>`.