In the k$-$nearest neighbors algorithm, increasing the value of k:

Group of answer choices

Does not affect model performance.

Always leads to overfitting.

Generally improves model performance by reducing bias but may increase variance.

May lead to underfitting.

Always improves model performance by reducing both bias and variance.