An autonomous research laboratory could conduct an entire study, from formulating questions to preparing manuscripts. Artificial intelligence (AI) offers one route towards this goal. Recent AI systems take on several research roles, but may draw conclusions that the evidence does not support. Here, we present an autonomous research laboratory in which agents based on large language models assess one another's work and can stop or redirect investigations. They can generate and reject research questions, require separate theoretical checks and conduct experiments designed to challenge their claims. We demonstrate these capabilities by investigating ensembles of quantum spins in diamond. The agents derived the minimum number of calibration settings needed to predict responses to a defined family of microwave pulse sequences. They checked the result computationally and designed experiments to distinguish competing predictions. An initial experiment agreed with the theoretical predictions, but repeated reference measurements showed that the signal was drifting during data collection. The agents therefore treated the result as inconclusive and revised the measurement protocol without relaxing the acceptance criteria. The revised experiment met those criteria and supported the theoretical predictions. This study demonstrates that collaborating AI agents can select research questions, derive theoretical results and test their predictions experimentally.
AI Summary
The paper asks whether collaborating LLM agents can formulate a theoretical research question, test it experimentally, and reject apparently supportive evidence when controls fail. A multi-agent workflow separates proposal generation, scientific assessment, computation, instrument operation and review. It is demonstrated on calibration of an inhomogeneous ensemble of NV-centre spins in diamond.
The central theoretical result is that $L^2+1$ linearly independent calibration settings are necessary and sufficient to predict mean responses to all phase-controlled sequences of at most $L$ pulses, without reconstructing the ensemble’s error distribution. This assumes fixed pulse duration, amplitude and shape, persistent spin-specific pulse errors, and a common fluorescence background and contrast. The result follows from the response-space dimension and a matching indistinguishability lower bound. Independent symbolic computations check dimensions 26 and 37 for $L=5,6$; certified numerical tests examine specified departures from persistent errors.
The experiment uses ten calibrations to predict four withheld three-pulse responses. An initial acquisition passed prediction-accuracy tests but failed drift-sensitive controls and was classified as inconclusive. After acquisition changes, a six-hour run passed the predefined criteria. The measured target contrast was $(40.0\pm0.4)\times10^{-3}$ weighted counts per shot, versus $(39.7\pm1.3)\times10^{-3}$ predicted; two specified interpolation baselines predicted zero.
Experimental support is limited to these targets and predefined tolerances, while robustness calculations cover specified perturbations rather than arbitrary noise. Humans assembled the apparatus and configured safety limits. The released workflow omits the study’s prompts; experimental data and analysis code are available on request.
Almost certainly largely AI-generatedAI use disclosedUnclear noveltyMissing comparison
The authors used generative AI agents for the research methodology, including scientific reasoning, computational analysis, experimental planning and execution, and manuscript drafting.
Central claim may not hold: The paper presents L²+1 calibration settings as a proved general minimum but omits the response-space dimension proof and the indistinguishable-ensemble lower-bound argument needed to establish that central theoretical claim.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
Comments
17 pages, 5 figures. Code available at https://github.com/E-zClap/auctopus
An autonomous research laboratory could conduct an entire study, from formulating questions to preparing manuscripts. Artificial intelligence (AI) offers one route towards this goal. Recent AI systems take on several research roles, but may draw conclusions that the evidence does not support. Here, we present an autonomous research laboratory in which agents based on large language models assess one another's work and can stop or redirect investigations. They can generate and reject research questions, require separate theoretical checks and conduct experiments designed to challenge their claims. We demonstrate these capabilities by investigating ensembles of quantum spins in diamond. The agents derived the minimum number of calibration settings needed to predict responses to a defined family of microwave pulse sequences. They checked the result computationally and designed experiments to distinguish competing predictions. An initial experiment agreed with the theoretical predictions, but repeated reference measurements showed that the signal was drifting during data collection. The agents therefore treated the result as inconclusive and revised the measurement protocol without relaxing the acceptance criteria. The revised experiment met those criteria and supported the theoretical predictions. This study demonstrates that collaborating AI agents can select research questions, derive theoretical results and test their predictions experimentally.
The paper asks whether collaborating LLM agents can formulate a theoretical research question, test it experimentally, and reject apparently supportive evidence when controls fail. A multi-agent workflow separates proposal generation, scientific assessment, computation, instrument operation and review. It is demonstrated on calibration of an inhomogeneous ensemble of NV-centre spins in diamond.
The central theoretical result is that L2+1 linearly independent calibration settings are necessary and sufficient to predict mean responses to all phase-controlled sequences of at most L pulses, without reconstructing the ensemble’s error distribution. This assumes fixed pulse duration, amplitude and shape, persistent spin-specific pulse errors, and a common fluorescence background and contrast. The result follows from the response-space dimension and a matching indistinguishability lower bound. Independent symbolic computations check dimensions 26 and 37 for L=5,6; certified numerical tests examine specified departures from persistent errors.
The experiment uses ten calibrations to predict four withheld three-pulse responses. An initial acquisition passed prediction-accuracy tests but failed drift-sensitive controls and was classified as inconclusive. After acquisition changes, a six-hour run passed the predefined criteria. The measured target contrast was (40.0±0.4)×10−3 weighted counts per shot, versus (39.7±1.3)×10−3 predicted; two specified interpolation baselines predicted zero.
Experimental support is limited to these targets and predefined tolerances, while robustness calculations cover specified perturbations rather than arbitrary noise. Humans assembled the apparatus and configured safety limits. The released workflow omits the study’s prompts; experimental data and analysis code are available on request.
The use of artificial intelligence (AI) in physics research is increasingly common, yet few guidelines exist for its effective use. Starting with a principles-based approach, we outline a set of ten requirements for collaborative AI agents in experimental physics, spanning scientific norms, effective agent use, and domain expertise. We take a human-centered approach in which agents act as collaborators that amplify rather than replace the expertise of a human researcher (user) who directs the work and is responsible for its results. We then present a framework designed to enforce these requirements, detailing methods for improving off-the-shelf behavior of large language models (LLMs) with custom scaffolding and describing general-purpose scientific workflows and skills. In particular, we highlight the power of closed-loop campaigns where agents work iteratively toward a well-defined goal. We demonstrate the potential of this mode of research through six case studies in experimental physics, covering device design, hardware control, neural network optimization, physical model identification, and data analysis. All six case studies result in useful work products that add utility to our research, including a neural network architecture and solid-state quantum defect photodynamic model that substantially outperform our published work; a microwave waveguide design that extends a published design concept to a fabrication-ready device meeting our experimental constraints; a quantum control pulse sequence candidate that improves simulated performance by more than an order of magnitude; commissioning of an NMR spectrometer to $\sim$1~Hz frequency precision on live hardware; and finally a data-fitting pipeline that achieves, to our knowledge, the first reconstruction of the Ising microstate of an artificial spin-ice array from diffraction-limited magnetic imaging.
We implement an agentic AI workflow built around a large language model (LLM) agent for autonomous experiments with nitrogen-vacancy (NV) centers in diamond. NV centers are a widely used platform for quantum sensing, and the ability to control many measurements from a computer makes NV experiments a natural setting for autonomous workflows. We make two main contributions. First, we demonstrate an autonomous NV experiment workflow that combines persistent project records, quantitative calculation and data analysis tools, and deterministic experiment control. In one autonomous experiment, the agent selected a single NV center, calibrated its resonant frequency, measured T2∗ with Ramsey measurements, and added a Carr--Purcell--Meiboom--Gill (CPMG) measurement to check a weak feature that could be related to nearby 13C. Second, we introduce two offline benchmarks that evaluate the agent's reasoning separately from laboratory execution. We evaluated both benchmarks with GPT-5.4, GPT-5.5, and GPT-5.6 Sol. In the Ramsey checkpoint benchmark, greater reasoning effort generally improved recognition of a residual resonance calibration offset. By contrast, in the pulsed optically detected magnetic resonance (pODMR) data evaluation benchmark, pulse sequence information alone produced more false positive resonance judgments at higher reasoning effort. Requiring an expected signal calculation kept false positive rates low across all three models and reasoning settings. The results suggest a clear division of labor for autonomous experiments. The agent forms scientific hypotheses and uses quantitative tools to evaluate data, while deterministic code controls the hardware and enforces safety constraints.
Artificial intelligence (AI) is used in numerous fields of science, yet the initial research questions and targets are still almost always provided by human researchers. AI-generated creative ideas in science are rare and often vague, so that it remains a human task to execute them. Automating idea generation and implementation in one coherent system would significantly shift the role of humans in the scientific process. Here we present AI-Mandel, an LLM agent that can generate and implement ideas in quantum physics. AI-Mandel formulates ideas from the literature and uses a domain-specific AI tool to turn them into concrete experiment designs that can readily be implemented in laboratories. The generated ideas by AI-Mandel are often scientifically interesting - for two of them we have already written independent scientific follow-up papers. The ideas include new variations of quantum teleportation, primitives of quantum networks in indefinite causal orders, and new concepts of geometric phases based on closed loops of quantum information transfer. AI-Mandel is a prototypical demonstration of an AI physicist that can generate and implement concrete, actionable ideas. Building such a system is not only useful to accelerate science, but it also reveals concrete open challenges on the path to human-level artificial scientists.
Large language model agents have begun to carry out laboratory measurements, but their scientific conclusions are useful only if they hold up under scrutiny. Demonstrating this requires controlled studies because models are stochastic and the context a model is given steers its conclusions. Here we introduce HERA, an autonomous agent designed to carry out cryogenic transport measurements of quantum devices. HERA draws on a knowledge base curated by lab members that holds measurement protocols, software tools, and literature references. We assessed HERA with a blinded trial against an expert on the stacking order of two multilayer graphene devices, an ablation of its knowledge base, and hundreds of offline replays in which models designed the trial's central measurement without running it. HERA's conclusions matched the expert's sealed conclusions on both tested devices. The ablation and the replay studies measured the impact of the knowledge base across different models, showing that with it every model designed better measurements and reasoned more soundly. To test how readily HERA extends to a new problem, we tasked it with finding and characterizing superconductivity in a different material system. This required a few hours of expert time to develop a new experimental protocol, after which HERA did both autonomously. These results show that, with a knowledge base designed for the task, language model agents are capable of sound experimental science.
With the rapid advancement of artificial intelligence (AI), a new paradigm is evolving that includes agentic AI systems capable of autonomously reasoning, planning, and executing multi-step actions. This chapter provides a detailed overview of agentic AI, AI agents, and the developing area of quantum AI agents – systems that utilize the capabilities of quantum computers to exceed the limits imposed by classical computer architecture. In addition to analyzing the architectural foundations of contemporary AI agents such as large language model (LLM)-based agents, multi-agent systems, and hybrid classical–quantum architectures, this chapter presents a unified theoretical framework for understanding how the application of quantum enhancement could result in exponential improvements in optimization, simulation, and pattern recognition abilities of agents using recent advances in quantum machine learning, variational quantum algorithms, and quantum neural networks. Furthermore, this chapter examines transformative applications across healthcare, finance, scientific discovery, and autonomous systems while identifying the current limitations of quantum AI, including decoherence, qubit counts, and alignment challenges. Finally, this chapter concludes that the intersection of agentic AI and quantum computing is likely one of the most significant technological developments of the 21st century, requiring sustained research, ethical frameworks, and inter-disciplinary cooperation.
Abstract: This study investigates an explainable quantum-logical multi-agent system (QL-MAS) for autonomous scientific discovery in high-dimensional research landscapes. The central methodological problem is that modern discovery settings combine large numbers of variables, heterogeneous evidence types, and partially conflicting reasoning signals, yet conventional single-model pipelines often compress this complexity into opaque rankings. The proposed framework addresses this limitation by treating candidate scientific hypotheses as quantum-logical superpositions of evidence generated by three complementary agents: an evidence agent based on mutual information, a structure agent based on principal-component loadings, and a rule agent based on interpretable decision structures. The resulting architecture does not claim physical quantum computation; rather, it uses quantum-logical amplitude composition as a mathematically disciplined way to represent concurrent support, interference, and consensus across agents. Empirical evaluation is conducted on three real public datasets that approximate heterogeneous scientific discovery landscapes: the Breast Cancer Wisconsin (Diagnostic) dataset (569 observations, 30 variables), the Wine Recognition dataset (178 observations, 13 variables), and the Diabetes dataset used in the least-angle-regression literature (442 observations, 10 baseline variables). These corpora were selected because they are non-synthetic, scientifically grounded, and sufficiently high-dimensional to test feature prioritization, interpretability, and cross-domain transfer. Across the three datasets, the proposed system identifies compact sets of high-salience hypotheses while preserving strong downstream predictive utility. The top-five features selected by QL-MAS attain cross-validated scores of 0.965 on breast cancer classification, 0.955 on wine classification, and 0.465 R² on diabetes progression, remaining close to full-feature baselines while offering substantially more interpretable evidence paths. The results indicate that explainable multi-agent discovery is strongest when high predictive relevance and rule-based consistency converge, and weakest when latent structure is diffuse or poorly aligned with supervised objectives. The analysis therefore supports the conclusion that autonomous discovery systems in research environments should not be assessed solely by predictive accuracy, but by the transparency of their evidential decomposition and by the stability of agreement across reasoning agents. The paper contributes a theoretically coherent framework, a real-data evaluation protocol, and a discussion of limitations, governance, and future directions for autonomous scientific reasoning systems. Keywords: autonomous scientific discovery; explainable AI; multi-agent systems; quantum logic; high-dimensional data; interpretable machine learning; research landscapes