NeurIPS scoring system is inherently subjective. People will have wildly different interpretations of, say, 3 vs 4, or 4 vs 5. You can get lucky and draw only r

NeurIPS scoring system is inherently subjective. People will have wildly different interpretations of, say, 3 vs 4, or 4 vs 5. You can get lucky and draw only reviewers that, on average, "overrate" papers in their batch. The opposite can happen too, obviously. 4444 vs 3444 is just noise.