Wrong Eigen Decomposition From Armadillo C++

Question

I'm relatively new to Armadillo and I'm trying to calculate the shared eigenvectors of two dense Matrix. I can't seem to get the correct values compared to python or a simple eigenvector calculator online. I know the python version of this code is correct but I don't know why armadillo isn't working in c++. As a simple test case I've used the code below

    mat B  = { {1, 2}, {3, 4} };
    mat C  = { {5, 6}, {0, 3} };

    cx_mat col_vec;
    cx_vec eig;

    eig_pair( eig, col_vec, B, C);
    col_vec.print("eigenvector");

Armadillo Result:
    eigenvector
    (-1.000e+00,+0.000e+00)    (+7.653e-01,+0.000e+00)
    (+6.377e-01,+0.000e+00)    (-1.000e+00,+0.000e+00)

If I compute the eigenvector in python I get a completely different result as seen below

   H_2 = np.array([[1, 2],[3, 4]])
   S_2 = np.array([[5, 6],[0, 3]])

   eigval, eig_vec = sp.linalg.eigh(H_2, S_2)
   print(eig_vec)

Python Result:
   [[-0.39880251  0.20237727]
   [ 0.26126793  0.51485182]]

I have no idea why this is the case and would really appreciate some help in calculating the correct eigenvalues in armadillo. If it's important at all I'm using WSL Ubuntu.