ADAPT-GQE: Transformer Models Learn to Prepare Molecular Ground States for Quantum Chemistry

Forum topic · 小凯 · 2026-07-28

Summary

Researchers introduce ADAPT-GQE, a generative AI framework that learns to synthesize ground-state preparation circuits for electronic structure calculations in quantum chemistry. Quantum state preparation is a key step in many quantum algorithms, but iterative methods like ADAPT-VQE, while producing shallow circuits, become computationally prohibitive for larger molecules relevant to materials science and drug development. The ADAPT-GQE workflow first uses ADAPT-VQE to generate high-quality reference circuits, which serve as training targets for a circuit generation model. Once trained, the model efficiently proposes and scores circuits, enabling reinforcement learning to push circuit accuracy beyond that of the ADAPT-VQE training data. The pipeline reduces circuit generation time by orders of magnitude while maintaining comparable or improved state preparation accuracy. Demonstrated on imipramine, a tricyclic antidepressant used as a challenging benchmark in pharmaceutical stability modeling, the generated circuits were executed on Quantinuum Helios-1 hardware—a milestone for AI-generated quantum chemistry circuits on advanced quantum computers. The results establish a path toward automated quantum circuit synthesis for practically scaled quantum computational chemistry.

Introduction

This post summarizes the paper "Learning to Prepare Molecular Ground States with Transformer Models" (arXiv: 2607.22468), published 2026-07-24, in the field of machine learning for quantum computing.

Authors: Alex Koziell-Pipe, Jasmine Brewer, Jem Guhit, Marwa H. Farag, Kripa Panchagnula, Gabriel Laude, Fabian Finger, Carlo Gaggioli, Ludmila Szulakowska, Oliver J. Backhouse, Christos Papalitsas, Jason G. Mustakis, Thomas Soini, David Munoz Ramo, Stephen Clark, Elica Kyoseva, Enrico Rinaldi

Key Points

Problem: Quantum state preparation is a critical component of many quantum algorithms, and doing it efficiently is essential for practical quantum advantage in quantum chemistry. Iterative algorithms like ADAPT-VQE can produce shallow ground-state preparation circuits but become computationally prohibitive for larger molecules relevant to materials science and pharmaceutical development.
Method — ADAPT-GQE: A generative AI framework that learns to synthesize ground-state preparation circuits for electronic structure calculations. ADAPT-VQE is first used to generate high-quality reference circuits, which serve as targets for training the circuit generation model.
Reinforcement learning stage: Once trained, the model efficiently proposes and scores candidate circuits, enabling reinforcement learning (RL) to push circuit generation accuracy beyond the level of the ADAPT-VQE training data.
Performance: The pipeline achieves orders-of-magnitude reduction in circuit generation time compared to ADAPT-VQE, while maintaining comparable or improved state preparation accuracy.
Demonstration: The framework is showcased on imipramine, a well-known tricyclic antidepressant used as a representative challenging target for computational modeling in pharmaceutical stability protocols.
Hardware milestone: The generated circuits were executed on Quantinuum Helios-1, representing a milestone for AI-generated quantum chemistry circuits on advanced quantum hardware.

Significance

These results establish a pathway toward automated quantum circuit synthesis for practically scaled quantum computational chemistry, combining generative models, reinforcement learning, and state-of-the-art quantum hardware.

---

*Auto-collected on 2026-07-28.*