It is important that code reviews are public enough that people see other people review each other's code - a system where the only code reviews you see are the ones you do and the ones you get leads to poor expectations of what they are for.

Having at least two very accomplished and (culturally and organisationally senior) people routinely "model" reasonable review behaviour can be stunningly effective.

Also this is an area where frequent team conversations about what good code is outside a review situation helps to build a certain culture. It helps step away from nitpicking and arguing.