In an attempt to better understand why the inverse Alexander polynomial appears in the MMR conjecture, I was reading the paper  of Bar-Natan and Garoufalidis giving their proof of the conjecture using weight systems. In particular, they discuss Rozansky's path-integral "proof" of the conjecture in section 1.4. I mostly follow this argument, except for the following claim:
Cheeger and Mueller proved that the Ray-Singer torsion is equal to the Reidemeister torsion, which by Milnor and Turaev was shown to be proportional to the inverse of the Alexander polynomial $A(K)$ of $K$.
My understanding was that Milnor and Turaev show that the torsion is proportional to the Alexander polynomial, not its inverse. Am I wrong? Is it a different torsion than the one I'm thinking of?
The torsion involves an alternating product, so it could just be different conventions. In that case understanding why this convention is right probably means digging deeper into the proof.