Sentence-transformers/all-mpnet-base-v2 requires Input Text after Cleaning or Raw Text Only

Hi Everyone,
I am using sentence-transformers/all-mpnet-base-v2 for measuring similarity between two paragraphs. The text I am giving to sentence-transformers/all-mpnet-base-v2 is cleaned text, and results are good, But as it compares sentences shouldn’t we give models raw text?

And also if someone have idea about best practices while using pretrained sentence-transformers/all-mpnet-base-v2. Kindly help me out

Ahsan Aftab - ML Engineer