This could be one of the most exciting result in a generation that has come out about the foundations of quantum mechanics. I have just downloaded the article and started to wade through it. In the meanwhile here is what Nature News has to say about it:
Their theorem effectively says that individual quantum systems must “know” exactly what state they have been prepared in, or the results of measurements on them would lead to results at odds with quantum mechanics. They declined to comment while their preprint is undergoing the journal-submission process, but say in their paper that their finding is similar to the notion that an individual coin being flipped in a biased way — for example, so that it comes up 'heads' six out of ten times — has the intrinsic, physical property of being biased, in contrast to the idea that the bias is simply a statistical property of many coin-flip outcomes.
I can't wait to get back from Ottawa to talk to my colleagues about this. Here is the link to the preprint of the paper by Matthew Pusey, Jonathan Barrett, and Terry Rudolph.