Overloaded unification

Hindley-Milner type inference is an application of unification,
for which the Martelli-Montanari algorithm is considered to be
at least as good as any other.  None of the older algorithms
allow overloading.

While I found some papers regarding algorithms for Haskell
type class overloaded unification, these won't do for my
purposes: type classes are too restrictive.  I haven't found
any papers regarding algorithms for unification in the face
of overloading.  (Maybe I am not using the correct search
terms.)

Modifying the Martelli-Montanari algorithm to treat
overloading does not seem difficult.  Nonetheless, if
somebody has already done it, I would prefer to read and
cite the literature.

Suggestions of papers (or even search terms) will be
appreciated.