Why does installing imbalanced-learn library work if it's done via Conda but not if it's via Pip3? Why are Conda and Pip different for this?

Question

I'm working in Jupyter Notebook and using the following versions:

Python 3.11.5, pip 23.2.1, conda 23.7.4, scikit-learn 1.3.0 during both pip3 and conda installations, and imbalanced-learn 0.11.0 in installations via both pip3 and conda as well.

So, after successfully installing imbalanced-learn library via Pip3,

pip install imblearn

when I attempted to import SMOTE algorithm from imblearn.over_sampling module in my Python notebook,

from imblearn.over_sampling import SMOTE

I unexpectedly encountered an Import Error, stating that:

ImportError: cannot import name '_MissingValues' from 'sklearn.utils._param_validation

Fortunately, subsequently un-installing imbalanced-learn and then proceeding to install it again, using Conda this time, resulted in no error from there onwards:

1. pip uninstall imblearn --yes

2. conda install -c conda-forge imbalanced-learn

So, after encountering and then resolving this issue, I tried finding out the reason why this happens in the first place. Unfortunately, I found nothing concrete behind this discrepancy. So, why are Conda and Pip different for this?

Thanks in advance.