Why do np.dot and A.dot produce different results when using a sparse matrix?

Question

The documentation for np.dot and .dot seem almost identical. However when running code with scipy sparse matrices I noticed they produce different outputs. I had assumed both me