I don't doubt your experience. Mine is totally different. I've never flipped one reviewer even thought I've always been polite and the reviewer clearly and objectively wrong. They just don't care. They have too much power. The AC just goes with the flow. This is not peer review, it's a joke.

This is not the case in ML conferences, according to my experience, it's a CV thing...

Well the rating could be hidden... Not that this is the problem, the problem is that the reviews are really entitled and not willing to stand corrected. I've these so many times. And I'd bet this kind of reveiwrs are phd students, not very clever ones.

I might be wrong of course, that's my impression from a discussion earlier today. BTW I think it'd be more fair not to know.

They won't know about the accept afaik.

In me experience this is absolutely rare. They don't flip.

I don't have an answer to your question, but given that the reviewers don't know the scores of the rest of the reviewers during the rebuttal, a good rebuttal could raise potentially all of them, but even with 2 borderlines being raised to weak accepts, accept should be possible. Good lucks folks.

Should be. Fightable, you should do a good rebuttal. Only hope the reviewers are not the classic cvpr reviewers (ignorant juniors)...