This paper is widely reported as saying that student evaluations anti-correlate with performance in later classes. I haven't read the paper, but I think that might be oversimplify the claim.

You might expect this result, if popular teachers are easy and don't push the students, but that's definitely not what's happening in this military academy with a uniform curriculum. But if what's popularly perceived as the dominant force in (american) evaluations has been eliminated, it's not clear whether this tells us much (about other american schools).