Update README.md

README.md

@@ -36,7 +36,7 @@ Reason for creating this model is related to experiment if there would be possib
 
 - **Model type:** T5
 - **Language(s) (NLP):** Estonian
-- **Finetuned from model:** [agemagician/mlong-t5-tglobal-base](https://huggingface.co/agemagician/mlong-t5-tglobal-base)
+- **Finetuned from model:** [agemagician/mlong-t5-tglobal-base](https://huggingface.co/agemagician/mlong-t5-tglobal-base). Vocabulary of the original model was reduced to keep only tokens present in training data.
 
 
 ## Uses
@@ -53,7 +53,7 @@ Biases coming from the original pre-trained model and from Estonian Parliament d
 
 ### Recommendations
 
-Don't use model in case you need very accurate results, model might miss important aspects from the original text.
+Don't use model in case you need very accurate results, model might miss important aspects from the original text and hallucinate.
 
 ## How to Get Started with the Model