← All research questions
Q15 / Preparation & verification / question

What is the STP cost of preparing a Quantum Max-Cut witness?

Compare verification by a weighted random-edge singlet test with STP preparation of a state whose Quantum Max-Cut objective is near optimal.

Discuss this question
Working mathematical formulation

For nonnegative rational edge weights wew_e, let W=∑ewe>0W=\sum_e w_e>0 and A=∑ewePs(e).A=\sum_e w_e P_s^{(e)}. A verifier receives one NN-qubit witness, draws edge ee with probability we/Ww_e/W, and accepts its singlet outcome; its acceptance is Tr⁡(Aρ)/W\operatorname{Tr}(A\rho)/W. With thresholds a>ba>b separated by 1/poly⁡(N)1/\operatorname{poly}(N), distinguish max⁡ρTr⁡(Aρ)/W≥a\max_\rho\operatorname{Tr}(A\rho)/W\ge a from max⁡ρTr⁡(Aρ)/W≤b\max_\rho\operatorname{Tr}(A\rho)/W\le b. Ask for the cost of preparing a near-optimal witness by STP, and compare the spin structure needed for preparation with what this verifier accesses.

Scope & assumptions

One randomized edge test defines an acceptance probability; estimating that probability empirically needs independent repetitions. A single unknown witness cannot be treated as arbitrarily many independent copies. The QMA\mathsf{QMA} verification task and a state-generation task have different input/output requirements.

Why this is interesting

The verifier accesses a weighted singlet expectation through local tests. The preparation problem asks how to achieve a near-optimal expectation across the interaction graph.

A place to start

Choose a small frustrated graph and compare an STP preparation protocol with randomized edge verification, including sampling error and back-action.

Related work

These works provide context and established results. They need not be the origin of this exact formulation.

Discuss this question

Reading is public. Posting requires a GitHub account. You can also write to Kunal.