Suppose a very large group of people who form their scientific beliefs at random. And then aliens kill everybody whose scientific beliefs aren't largely true. Surely the survivors don't know their scientific beliefs to be true, assuming they don't know about the aliens killing off those who were wrong. After all, they didn't know their beliefs prior to the massacre, and the massacre of the erring didn't give them knowledge. (This case may be a problem for certain kinds of reliabilism.)
Suppose now that we came to believe certain truths--whether mathematical or empirical or moral--"for evolutionary reasons". In other words, those who had these beliefs survived and reproduced, and those who didn't didn't. Let's even suppose that there is a tight connection between the truth of the beliefs and their fitness value. Nonetheless I am not sure that this story is sufficiently different from the story about aliens to make the beliefs into knowledge. The crucial difference, I guess, is that the story about the aliens is a single-generation story, while the evolutionary story is a multi-generation story. But I am not sure that matters at all. Suppose, for instance, that we modify the alien story to add a new generation who takes their scientific beliefs from the survivors of the first generation. Surely if you get your scientific beliefs from people who don't know, and no additional evidence is injected, you don't know either.