Update README.md

README.md

@@ -158,4 +158,27 @@ On the 35th epoch, the model achieved the following loss:
 | 35.40     | 58.55	       | 39.46 | 52.43 |
 
 While a SacreBLEU score of 35 is a moderate score, it is important to consider that docstrings in Google style format vary extremely. Some are outliers that have extra sections
-that is usually not included in the general population which leads the model to generate "hallucinations".
+that is usually not included in the general population which leads the model to generate "hallucinations". An example of this is this particular sample:
+
+```
+Reference:  Validate timestamp specified by request.
+
+See `validate.request` for additional info.
+
+Args:
+stamp: str. Time request was made as ISO 8601 timestamp.
+tolerance: int. Number of seconds request remains valid from timestamp.
+
+Returns
+bool: True if valid, False otherwise.
+-----------------------------------------------------------------------
+Prediction:  Validate timestamp.
+
+Args:
+stamp (str): A date string in the format YYYY-MM-DDThh:mm:ss.######[+-]##:##
+
+Returns:
+bool: True if valid, False otherwise.
+```
+
+As you can see, the model generated gibberish in the prediction's Args section specifically the string format for the date.