Why bert score is high even without semantically similar sentences?

[dhandhalyabhavik]

Pred: The cat sat on the mat
Ref: A car was sold yesterday.

Prediction 1:
Precision: 0.896
Recall: 0.893
F1: 0.894

[Skripkon]

Hey, I think the code snippet below answers your question:

The full code can be found here: --> link

[Skripkon]

For more on how this baseline rescaling works, you can read about it at the very top of page 5 of the official article