From the abstract of the original article (emphasis mine):
Measuring a quantum system can randomly perturb its state. The strength and nature of this back-action depend on the quantity that is measured. In a partial measurement performed by an ideal apparatus, quantum physics predicts that the system remains in a pure state whose evolution can be tracked perfectly from the measurement record. We demonstrated this property using a superconducting qubit dispersively coupled to a cavity traversed by a microwave signal. The back-action on the qubit state of a single measurement of both signal quadratures was observed and shown to produce a stochastic operation whose action is determined by the measurement result. This accurate monitoring of a qubit state is an essential prerequisite for measurement-based feedback control of quantum systems.
That is, the abstract itself describes the fact that the measurement changes the state; what they are saying is that in their apparatus, how the state changes upon measurement of a subsystem can be reliably inferred from the measurement outcome, as in theoretical treatments of wave-function collapse of entangled states.