Drop in r2 score when trying out Xgboost in different versions of python

Question

I have been trying to run xgboost model on the same training data in different versions of python. One version is python 3.6 and the other is python 3.10. But in both the model I can see a significant drop in R2 score.

• In python 3.6 - 0.33 r2 score
• In python 3.10 - 0.075 r2 score.

What could be the potential reason for this difference given we use the same training data.

Thanks in advance.

the same training data, generated a pickle file in both versions and did scoring.