Action-Factored Multi-Agent Reinforcement Learning for Scalable Quantum Device Tuning

2026-07-10

AI Summary

This work addresses the instability in tuning multi-quantum-dot arrays caused by strong parameter crosstalk and non-stationary environmental conditions by proposing QADAPT, a scalable multi-agent reinforcement learning framework based on action space factorization. By decoupling the action space to mitigate inter-agent interference and integrating local observations with shared reward-based policy learning, QADAPT enables efficient cooperative tuning. Its key innovation lies in an action factorization mechanism that facilitates zero-shot generalization to quantum devices of unseen scales, with convergence steps remaining nearly constant regardless of system size. Experimental results demonstrate that QADAPT reliably and rapidly converges to target operating regimes across quantum devices of varying scales, offering an efficient and scalable solution for calibrating large-scale quantum processors.

Abstract

Cooperative multi-agent reinforcement learning is well suited to problems with large parameter spaces and exploitable local structure, such as the tuning of electrostatically-defined quantum-dot arrays. However, if parameter cross-talk is strong, a non-stationary environment from the perspective of any individual agent can destabilize learning - the same effect that plagues manual tuning of such systems. We propose using a factored representation of the action space, learned online, to decouple agents and minimize their interference. Our framework, QADAPT, uses this factorization to efficiently learn shared policies based on local measurements and rewards. With this modular strategy, we achieve zero-shot generalization to unseen quantum device sizes and maintain an approximately constant number of convergence steps to reach target regimes. This work provides a scalable route toward the rapid calibration of large-scale quantum processors.

Problem

• multi-agent reinforcement learning
• quantum device tuning
• parameter cross-talk
• non-stationary environment
• scalable calibration

Innovation

• action-factored representation
• multi-agent reinforcement learning
• quantum device tuning
• zero-shot generalization
• scalable calibration