Sentence length influence on similarity

Hi everyone,
I’m using sentence transformers to compute similarity between text. (embedding, mean pooling, cosine similarity)
I get the impressions that beyond text meaning, text length plays a huge role in similarity.
As anyone experienced the same thing ?

I fine tuned a model to compute similarity between names. This is a toy example:

 name0 name1 label
 Test  Test  y
 Test  Hi    n

I fined-tuned a model using the label and feeding it with pairs of names.

I also found out that longer names tend to be predicted as y.