p=1 QAOA MaxCut Ansatz on a Triangle (K3) - references arXiv:2607.20225

This card builds a p = 1 QAOA ansatz - the textbook Quantum Approximate Optimization Algorithm circuit - for MaxCut on a three-vertex triangle (K3). Starting from an equal superposition (a Hadamard on each qubit), it applies one cost layer that entangles each graph edge via a CX-RZ-CX ZZ-interaction, followed by one transverse-field mixer layer of RX rotations. The cost and mixer angles here are fixed, representative values chosen only to produce a concrete deterministic circuit; they are not claimed to be optimal.

What the referenced paper reports: Kim and co-authors introduce DQAOA-GPT, a hybrid framework that pairs the distributed QAOA - which decomposes a large optimization problem into smaller sub-problems - with a GPT-based generative model that directly produces quantum circuits for those sub-problems, replacing iterative variational parameter updates. Their abstract benchmarks DQAOA-GPT against conventional DQAOA on dense HUBO problems with up to 100 decision variables, reporting reduced computational cost while maintaining competitive solution quality, with larger speed-ups for larger sub-problems. They position it as a foundation for larger-scale optimization in hybrid HPC-QC environments.

The circuit on this Provenova card is inspired by that line of work only in that it exercises the same QAOA-family primitive at textbook scale. It is a small, self-contained p = 1 QAOA MaxCut ansatz, not the paper's distributed decomposition, its GPT-generated circuits, or its HUBO benchmarks. This is a deterministic simulator run on Provenova and not a reproduction of the paper's methods or reported results, and no endorsement by the authors is implied. See the linked arXiv abstract for the framework and benchmarks.