This thesis identifies a formal connection between physical problems related to entanglement detection and complexity classes in theoretical computer science. In particular, we prove that to nearly every quantum interactive proof complexity class (including BQP, QMA, QMA(2), QSZK, and QIP), there corresponds a natural entanglement or correlation detection problem that is complete for that class. In this sense, we can say that an entanglement or correlation detection problem captures the expressive power of each quantum interactive proof complexity class, and the contrast between such problems gives intuition to the differences between classes of quantum interactive proofs. It is shown that the difficulty of entanglement detection also depen...