Automated Discovery of Non-Standard Quantum Gate Decompositions with AI Research Agents

Christopher Poon, Zhiyuan Wang, Xiao-Yang Liu

Keywords: AI Agents, Automated Discovery, Quantum Computing, Circuit Decomposition

TL;DR: An AI agent automatically discovers and verifies hardware-efficient decompositions of fundamental quantum gates.

Abstract: The automated design of efficient quantum circuits is a critical challenge in the Noisy Intermediate-Scale Quantum (NISQ) era. This paper demonstrates a novel methodology where an AI research agent is tasked with discovering non-standard decompositions for fundamental quantum gates. By providing the agent with a target unitary and a set of structural constraints, we successfully generated and rigorously verified alternative circuits for the Toffoli, SWAP, and iSWAP gates. While these decompositions are not entirely new algebraic identities, they represent systematically rediscovered and structurally distinct implementations that highlight different hardware-relevant trade-offs. The results showcase reductions in circuit depth or the elimination of costly non-native gates, underscoring the potential of AI-driven workflows not only in quantum circuit optimization but also as a methodological advance toward automated, verifiable scientific discovery.

Submission Number: 270