Typing of the `__get_item__` method of numpy array subclasses

Let's consider a subclass of numpy's ndarray class:

import numpy as npclass ArraySubClass(np.ndarray):    def __new__(cls, input_array: np.ndarray):        obj = np.asarray(input_array).view(cls)        return obj

Then, taking a slice of an ArraySubClass return an object of the same type, as explained in the documentation:

>>> type(ArraySubClass(np.zeros((3, 3)))[:, 0])<class '__main__.ArraySubClass'>

So far so good, but I start getting unexpected behavior when I use the static type checked pyright, as seen with the example below:

def f(x: ArraySubClass):    print(x)f(ArraySubClass(np.zeros((3, 3)))[:, 0])

The last line raises an error from pyright:

 Argument of type "ndarray[Any, Unknown]" cannot be assigned to parameter "x" of type "ArraySubClass" in function "f"  "ndarray[Any, Unknown]" is incompatible with "ArraySubClass"

What causes this behavior? Is this a pyright bug? Is the signature of the __get_item__ method from np.ndarray given by the type hints incorrect? Or perhaps should I override this method in ArraySubClass with the correct signature?