Quantum planted-clique detection in the conjectured classical-hard regime
Determine whether a uniform polynomial-time quantum algorithm can achieve constant-advantage planted-clique detection from one classical graph when k=floor(n^(1/2-epsilon)) for any fixed 0<epsilon<1/2.
References
A uniform polynomial-time quantum algorithm that achieves constant-advantage detection at $k=\lfloor n{1/2-\varepsilon}\rfloor$, for even one fixed $0<\varepsilon<1/2$, would establish an average-case quantum advantage under the classical conjecture. Here both algorithms receive the same input: one classical graph. Whether such a quantum algorithm exists remains open.
An efficient measurement with constant advantage on these reduced states would give an efficient planted-clique distinguisher. Our contribution is to formulate this measurement problem explicitly and establish its statistical sufficiency; its computational complexity remains open.
We determine how many copies of the state are necessary and sufficient to discriminate the signal. We prove a lower and upper bounds for joint measurements. We then identify the limiting experiment as observing the graph up to complementation and derive its Helstrom and pretty good decision rules. These results quantify the information in the encoding; efficient detection in the conjectured hard regime remains open.
The computational question is whether a smaller, efficiently implementable action can expose a useful part of this separation through its labels.