Progress in quantum hardware requires a growing volume of experiments across diverse devices, straining the capacity of human experts. Automation remains challenging because device differences and drift can render workflows developed on one processor unsuitable for another. Here we present an agent-assisted experimental strategy for conducting experiments across three-qubit nuclear magnetic resonance processors. Developed on one processor and applied sequentially to four others, the strategy uses an agent to assess experimental evidence, select subsequent measurements and adapt procedures, with operators providing instrument preparation and physical intervention. Using three-qubit Grover state preparation as a common task, each deployment combined device-specific calibration, control-pulse optimization and measurement. The experimental strategy was refined through sequential deployment, and all four processors ultimately produced outputs dominated by their respective marked states. The marked-state populations of the reconstructed dominant eigenstates ranged from \(0.910\) to \(0.959\), with relative deviations below \(3.8\%\) from the ideal value of 0.945. These results demonstrate that an agent-assisted experimental strategy can be reused and refined across processors to realize a common quantum task, providing a basis for scaling experimental practice beyond individual devices.
AI Summary
The paper asks whether an agent-assisted experimental strategy can realize a common multiqubit task on different processors by adapting calibration, control and measurement procedures. A language-model agent coordinates deterministic tools, while operators prepare and adjust instruments, restore servers and authorize interventions. The strategy is developed on one three-qubit liquid-state NMR processor and deployed sequentially to four others of the same family, sharing an experimental interface.
The task is state preparation corresponding to two Grover iterations, implemented with device-specific 11-ms GRAPE pulses. Each deployment rebuilds its Hamiltonian model and measurement references. Recorded adaptations include rejecting inherited calibration settings, reacquiring unstable references and revising spectral fitting for overlapping transitions. Seven-setting tomography gives terminal marked-state populations of $0.929$, $0.959$, $0.936$ and $0.910$, versus the ideal $121/128\approx0.945$, with population-distribution total-variation distances of $0.020$–$0.047$.
These headline values describe the dominant eigenvectors of reconstructed effective density matrices after numerical purification, not the measured mixed states or absolute molecular populations. Before purification, terminal populations are $0.860$, $0.959$, $0.826$ and $0.910$. Fifteen subsequent tomography repeats across five completed sessions show variable repeatability, assessed retrospectively without a numerical cutoff. The paper notes that purification affects both agreement and repeatability; transfer-context isolation was procedural rather than enforced; some agent configuration and entry-context records are incomplete; and simulated robustness uses the optimization ensemble itself. Transfer beyond this processor family and hardware-independent task generation are not evaluated.
Little sign of AI-generated textAI use disclosedIncrementalMissing comparisonUnclear novelty
The Q0 records identify the agent as Codex based on GPT-5, although the exact model build was unavailable.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
Progress in quantum hardware requires a growing volume of experiments across diverse devices, straining the capacity of human experts. Automation remains challenging because device differences and drift can render workflows developed on one processor unsuitable for another. Here we present an agent-assisted experimental strategy for conducting experiments across three-qubit nuclear magnetic resonance processors. Developed on one processor and applied sequentially to four others, the strategy uses an agent to assess experimental evidence, select subsequent measurements and adapt procedures, with operators providing instrument preparation and physical intervention. Using three-qubit Grover state preparation as a common task, each deployment combined device-specific calibration, control-pulse optimization and measurement. The experimental strategy was refined through sequential deployment, and all four processors ultimately produced outputs dominated by their respective marked states. The marked-state populations of the reconstructed dominant eigenstates ranged from (0.910) to (0.959), with relative deviations below (3.8%) from the ideal value of 0.945. These results demonstrate that an agent-assisted experimental strategy can be reused and refined across processors to realize a common quantum task, providing a basis for scaling experimental practice beyond individual devices.
The paper asks whether an agent-assisted experimental strategy can realize a common multiqubit task on different processors by adapting calibration, control and measurement procedures. A language-model agent coordinates deterministic tools, while operators prepare and adjust instruments, restore servers and authorize interventions. The strategy is developed on one three-qubit liquid-state NMR processor and deployed sequentially to four others of the same family, sharing an experimental interface.
The task is state preparation corresponding to two Grover iterations, implemented with device-specific 11-ms GRAPE pulses. Each deployment rebuilds its Hamiltonian model and measurement references. Recorded adaptations include rejecting inherited calibration settings, reacquiring unstable references and revising spectral fitting for overlapping transitions. Seven-setting tomography gives terminal marked-state populations of 0.929, 0.959, 0.936 and 0.910, versus the ideal 121/128≈0.945, with population-distribution total-variation distances of 0.020–0.047.
These headline values describe the dominant eigenvectors of reconstructed effective density matrices after numerical purification, not the measured mixed states or absolute molecular populations. Before purification, terminal populations are 0.860, 0.959, 0.826 and 0.910. Fifteen subsequent tomography repeats across five completed sessions show variable repeatability, assessed retrospectively without a numerical cutoff. The paper notes that purification affects both agreement and repeatability; transfer-context isolation was procedural rather than enforced; some agent configuration and entry-context records are incomplete; and simulated robustness uses the optimization ensemble itself. Transfer beyond this processor family and hardware-independent task generation are not evaluated.
Quantum computers are moving from research laboratories to industrial machines accessible via the cloud and integrated into high-performance computing facilities. However, translating theoretical quantum protocols into hardware experiments remains a major bottleneck, requiring expertise across protocol design, compilation, simulation, and cloud execution. Here, we introduce an agentic workflow that automates this pipeline on neutral-atom quantum processors (here two Pasqal QPUs available on the cloud) while keeping the researcher in the loop for critical validation. In three case studies from many-body physics and optimization, the agent went from published paper or patent to a QPU campaign run overnight. In particular, human intervention was crucial to ensure scientific validity: the agent selected an inadequate observable in one experiment and constructed a plausible but incorrect hardware diagnosis in another, with both failures detected only through domain-expert review. Finally, we use a second agent to classify a corpus of 633 Rydberg-array arXiv papers and show that nearly half are implementable on present-day QPUs while identifying specific hardware upgrades needed for the rest. Together, these results demonstrate that agentic workflows provide a practical bridge between theoretical ideas and physical hardware, opening quantum experimentation to a much broader scientific community.
We propose and experimentally demonstrate sequential quantum computing (SQC), a paradigm that utilizes multiple homogeneous or heterogeneous quantum processors in hybrid classical-quantum workflows. In this manner, we are able to overcome the limitations of each type of quantum computer by combining their complementary strengths. Current quantum devices, including analog quantum annealers and digital quantum processors, offer distinct advantages, yet face significant practical constraints when individually used. SQC addresses this by efficient inter-processor transfer of information through bias fields. Consequently, measurement outcomes from one quantum processor are encoded in the initial-state preparation of the subsequent quantum computer. We experimentally validate SQC by solving a combinatorial optimization problem with interactions up to three-body terms. A D-Wave quantum annealer utilizing 678 qubits approximately solves the problem, and an IBM's 156-qubit digital quantum processor subsequently refines the obtained solutions. This is possible via the digital introduction of non-stoquastic counterdiabatic terms unavailable to the analog quantum annealer. The experiment shows a substantial reduction in computational resources and improvement in the quality of the solution compared to the standalone operations of the individual quantum processors. These results highlight SQC as a powerful and versatile approach for addressing complex combinatorial optimization problems, with potential applications in quantum simulation of many-body systems, quantum chemistry, among others.
Superconducting quantum computing is one of the most mature solid-state platforms for quantum computation, with processors exceeding one hundred qubits. Yet further scaling toward fault-tolerant quantum computing is increasingly constrained by calibration complexity. Conventional scripts are brittle to anomalous signals, and expert judgment is bounded by cognitive bandwidth and serial operation time, failing to keep pace with system scale. Here we report Vibe Calibration, an autonomous calibration system orchestrated by large language model agents, which distills expert tacit knowledge into reusable Skills. Each Skill is organized as a decision tree that packages parameterized measurement commands, quantitative acceptance criteria, and audit records, enabling autonomous execution and self-healing. We capture this knowledge through a three-phase human-in-the-loop distillation process and fine-tune a large language model on validated trajectories. On a 112-qubit processor with frequency-tunable transmons, the system autonomously completes calibration of 108 out of 112 qubits in 4.7 hours, achieving a 4--5× speedup over manual calibration of the full 112 qubits. A cross-validated comparison with expert manual calibration on a 16-qubit subset shows agreement on 14 out of 16 qubits. More importantly, the model demonstrates transferable calibration workflows across devices. While low-level control scripts require minor interface adaptation for different hardware platforms, the core decision logic and task orchestration generalize to new processors, demonstrating a reusable laboratory interface rather than a memorized script.This work demonstrates, for the first time, fully autonomous calibration of a hundred-qubit superconducting processor through reusable and auditable Skills, removing a critical barrier to scalable quantum hardware operation.
With the development of quantum information and quantum computation science, quantum information processor has been widely used in different areas such as quantum simulation, quantum computation and quantum metrology and so on. To make quantum computer come true, we need to increase the number of controllable qubits of the system and improve the controllability to perform complex quantum manipulation. As a good experimental testbed for quantum information processing, nuclear magnetic resonance (NMR) spin system provides rich and sophisticated quantum control methods. In recent years a lot of multi-qubit experiments have been performed on the platform and a series of experimental technologies have been developed. In this paper, we firstly explain the difficulties of multi-qubit NMR experiments. Then by focusing on the experiment of 7-qubit labelled pseudo-pure state preparation and other relevant experiments, we review the technologies in multi-qubit experiments. Using the radio frequency selective method, the inhomogeneities of the radio frequency pulses are reduced and the spectral resolution is improved. After performing 1/2 spin selective sequence, we can regard the three methyl protons in the sample of crotonic acid as a single 1/2 spin nucleus and treat the whole molecule as a 7-qubit quantum information processor. We utilize Gauss pulses, Hermite pulses, composite pulses and gradient ascent pulse engineering (GRAPE) pulses to implement basic /2 and rotation operations. The GRAPE pulses are calculated by subspace GRAPE program to speed up the computation greatly. The errors of the basic pulses caused by chemical shift and J-coupling evolution can be estimated by the program of pulse compilation. It divides the errors of the pulses into a series of post-errors and pre-errors. A program of sequence compilation is used to eliminate the accumulated error of the whole pulse sequence, reduce the number of pulses and optimize the experimental duration. A variety of methods of quantum state tomography have been proposed to improve the efficiency of reading out information about quantum state. As an experimental example, we combine the above experimental technologies and perform the experiment of 7-qubit labelled pseudo-pure state preparation by using the method of cat state preparation. The sequence of cat state preparation consists of three steps:encoding procedure, phase cycling and decoding procedure. We use 14 experiments to realize the phase cycling and acquire the final 7-qubit labelled pseudo-pure state. The total duration of experimental sequence is about 132 ms. All the readout spectra have the similar shapes to the theoretical expectations. Finally we give an outlook for further research in this direction.
Measurement is an essential component of robust and practical quantum computation. For superconducting qubits, the measurement process involves the effective manipulation of the joint qubit-resonator dynamics, and it should ideally provide the highest quality for qubit state discrimination with the shortest readout pulse and resonator reset time. Here, we harness model-free reinforcement learning (RL), together with a tailored training environment, to achieve this multifaceted optimization task. Using the IBM quantum device, we demonstrate that the pulse obtained by the RL agent not only successfully achieves state-of-the-art performance, with an assignment error of ( 4.6 ± 0.4 ) × 10 − 3 , but also executes the readout and the subsequent resonator reset almost three times faster than the system’s default process. Furthermore, the learned waveforms are robust against realistic parameter drifts and follow an intuitive form, making them readily implementable on existing hardware with little computational overhead. Our results provide an effective readout strategy to boost the performance of superconducting quantum processors and demonstrate the value of RL in providing optimal and practical solutions for complex quantum information processing tasks.
As spin-based quantum systems scale, their setup and control complexity increase sharply. In semiconductor quantum dot (QD) experiments, device-to-device variability, heterogeneous control-electronics stacks, and differing operational modalities make it difficult to reuse characterization, calibration, and control logic across laboratories. We present FAlCon, an open-source software ecosystem for portable, automated characterization and tuning measurement workflows. FAlCon provides (i) a lightweight domain-specific language for expressing state-based tuning logic in a hardware-agnostic form; (ii) specialized transmittable libraries of physics-informed QD data structures (``tuning vernacula''); and (iii) extensible libraries of shared measurement protocols enabling an interoperable lab-agnostic measurement stack. By separating algorithm intent from instrument realization, while preserving traceability and supporting typed scripting, FAlCon enables researchers and engineers to exchange, adapt, and deploy characterization and autotuning routines across heterogeneous QD setups. The framework supports all users, ranging from end users running prebuilt algorithms with custom initial conditions to developers extending instrumentation support and contributing new tuning strategies. Although the present release targets QD experiments, other qubit modalities and scientific experiments could reuse FAlCon's modular abstractions by providing new tuning data types and instrument control templates.