Pith. sign in

REVIEW 4 major objections 7 minor 22 references

An LLM agent can run autonomous nitrogen-vacancy quantum sensing experiments when it forms hypotheses and checks data quantitatively, while deterministic code alone controls the hardware and safety.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-31 00:08 UTC pith:BCTYZQUY

load-bearing objection Real NV autonomy demo plus useful evidence that more LLM reasoning helps hypothesis integration but can hurt bare data calls unless you force a quantitative check. the 4 major comments →

arxiv 2607.25145 v1 pith:BCTYZQUY submitted 2026-07-27 quant-ph cs.AI

Agentic AI for Scientific Reasoning in Autonomous Quantum Sensing Experiments

Takuya Isogawa , Ryotaro Okabe , Nutdech Phadetsuwannukun , Mingda Li , Paola Cappellaro This is my paper
classification quant-ph cs.AI
keywords nitrogen-vacancy centersagentic AIautonomous experimentsquantum sensingLLM reasoningpODMRRamsey spectroscopydeterministic hardware control
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper shows that a large language model agent can drive multi-step nitrogen-vacancy (NV) center experiments in diamond without continuous specialist supervision. The agent keeps persistent project records, writes analysis and simulation scripts, chooses measurements such as pODMR, Ramsey, and even an unrequested CPMG follow-up, and updates scientific conclusions as data arrive. Hardware access is deliberately walled off: only verified deterministic software queues jobs, enforces limits, and runs instruments. Two offline benchmarks separate reasoning quality from lab execution. Higher reasoning effort generally helped the agent notice a residual resonance-calibration offset that had skewed Ramsey interpretation. By contrast, judging whether a pODMR trace contains a resonance from pulse-sequence information alone produced more false positives as reasoning increased, unless the agent was required to compute an expected signal first. The practical message is a division of labor: use the agent for hypothesis formation and quantitative evaluation, and keep control and safety in ordinary code.

Core claim

An agentic workflow that pairs an LLM for scientific orchestration with deterministic experiment control can complete autonomous NV characterization—selecting an aligned center, calibrating resonance, measuring T2*, and testing weak nearby-13C signatures—while offline benchmarks show that more reasoning helps multi-evidence hypothesis formation but can inflate false-positive resonance calls unless an expected-signal calculation is required.

What carries the argument

The split architecture: persistent project records plus quantitative Python tools on the agent side, and a shared-folder job queue with a deterministic verifier and no direct instrument access on the experiment side; evaluated with Ramsey checkpoint packages and a zero-shot pODMR resonance-judgment benchmark under three prompt conditions.

Load-bearing premise

That frozen offline checkpoints from one past project and orientation-based pODMR labels are a fair enough stand-in for live, multi-step laboratory judgment that the reasoning-effort trends should guide real deployments.

What would settle it

Rerun the same Ramsey checkpoint and pODMR suites with the same models and prompts: if higher reasoning no longer raises residual-calibration pass rates, or if requiring an expected-signal calculation no longer holds false-positive rates near zero across models, the division-of-labor claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Autonomous NV runs can recover from failed tracks and invalid acquisitions and still reach supported T2* and 13C-style conclusions.
  • Higher reasoning effort should be reserved for integrating context and forming hypotheses, not for routine presence/absence data calls.
  • Validated fitting, simulation, and classification routines should be exposed as tools the agent chooses rather than replacing the lab software stack.
  • Detailed agent records of measurements and decisions can be turned into deterministic procedures once they prove reliable.
  • The same split could extend commercial NV instruments toward unattended magnetic imaging, geology, or GPS-denied navigation tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Other computer-controlled quantum platforms (superconducting qubits, trapped ions, cold atoms) likely need the same hypothesis-versus-hardware split rather than end-to-end LLM control.
  • As models get better at long-context hypothesis formation, the binding constraint may shift from model IQ to context management and safety verifiers over multi-day runs.
  • A simple contrast-depth or expected-signal gate discovered in exploration could be promoted to a hard tool, cutting false positives without relying on reasoning effort.
  • Benchmarks built from projects where the agent originally missed a calibration offset are especially diagnostic for whether agents can catch their own systematic errors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The manuscript presents an agentic-AI workflow for autonomous NV-center quantum sensing experiments, in which an LLM agent maintains persistent project records, plans measurements, and writes its own analysis code, while deterministic software alone controls hardware and enforces safety. Three autonomous case studies are reported (one detailed: NV selection, resonance calibration, Ramsey T2* measurement, and an agent-initiated CPMG follow-up on a weak 13C-like feature; one required human advice to recognize a residual calibration offset). The second contribution is two offline benchmarks evaluated on three model versions (GPT-5.4, GPT-5.5, GPT-5.6 Sol) at four reasoning-effort settings: (i) a Ramsey checkpoint benchmark (5 checkpoints archived from the project that originally missed the calibration offset; 20 reps per cell; 1,200 runs; binary manual rubric requiring the response to link a residual Ramsey frequency offset to resonance/microwave calibration); and (ii) a pODMR resonance-judgment benchmark (96 labeled measurements, 3 prompt conditions, 3 replicates; 10,368 decisions; measurement-level bootstrap CIs). The headline findings are that higher reasoning effort generally raised Ramsey pass rates, while in the pODMR task higher reasoning effort increased false positives under sequence-only prompts, and requiring an explicit expected-signal calculation kept false positive rates at 0–3.7% across all models and settings.

Significance. If the results hold, this is a useful and unusually careful contribution to the growing literature on LLM agents for physics experiments. The division-of-labor conclusion — LLM for hypothesis formation and orchestration, deterministic code for hardware control and routine data judgment — is well motivated and directly actionable for the autonomous-experiment community. Particular strengths deserve emphasis: the full public release (project records, raw data, per-run benchmark outputs, binary scores with rationales, and figure code) makes the benchmark claims auditable and reproducible; the pODMR benchmark uses labels derived independently of model outputs (known NV orientation vs. scan range) with a pre-registered-style fixed rubric and measurement-level bootstrap confidence intervals; and the finding that increased reasoning effort can *increase* false-positive resonance judgments is counterintuitive, falsifiable, and of practical significance beyond NV centers. The case studies are documented at a level of detail (18.9 h run, full decision traces) that is rare in this literature. The work is methodologically sounder than most end-to-end agent demonstrations, though the Ramsey ben

major comments (4)
  1. [Results / Fig. 3b–e; Methods 'Ramsey Checkpoint Benchmark'] The Ramsey benchmark's effective sample size is much smaller than '400 runs per model' implies, and no uncertainty quantification is given for the central trend. The five checkpoints are sequential states of a single project sharing one causal chain (memory instruction → minimum-sampled-point frequency choice → subsequent Ramsey acquisitions), so the 20 replicates per cell measure stochastic model variation, not task diversity. The heat maps confirm checkpoint effects dominate: cp04 yields zero passes for all 36 model×reasoning cells, and GPT-5.4's passes are concentrated at cp01 (7/20 at low already). Pooling across checkpoints to report '7/100 → 20/100' (GPT-5.4) without confidence intervals or a trend test leaves it unclear whether the aggregate monotonicity is robust to checkpoint composition. Please (i) report per-cell CIs that treat checkpoints as clusters (or state clearly that wi
  2. [Methods 'Ramsey Checkpoint Benchmark'; Discussion ¶2] The benchmark is constructed from the single archived project that is known, in hindsight, to have failed in exactly the way the rubric tests. The pass criterion (response must link the residual Ramsey offset specifically to resonance/microwave calibration) encodes the known answer to that one failure. This is legitimate as a retrospective probe, and the prompts correctly exclude the later human advice, but it is a correctness risk for the generalization drawn in the Discussion: 'These gains indicate that LLM agents are becoming increasingly practical for experimental automation.' A single hindsight-selected failure mode cannot support a deployment-level claim about hypothesis formation in general. Please add an explicit limitations paragraph stating that the benchmark measures one induced error type in one project, and temper the Discussion sentence accordingly (e.g., restrict the claim
  3. [Methods 'Ramsey Checkpoint Benchmark'; Table 2] The 1,200 binary scores are manual, but the manuscript does not state who scored, whether the scorer was blinded to model and reasoning setting, or whether any inter-rater reliability check was performed. Table 2 shows the rubric requires judgment on borderline cases (e.g., cp01__high__rep07 fails despite mentioning 'residual frequency error' because no calibration cause is stated), so scoring is not purely mechanical and could correlate with the swept variables if unblinded. Since all run records are already public, a blinded second-scorer audit on a random subset (e.g., 200 runs stratified by model and reasoning level) with reported Cohen's κ is feasible and would substantially strengthen the benchmark. At minimum, the scoring protocol (scorer identity, blinding, order of scoring) must be stated.
  4. [Results Fig. 4; Supp. Note 3 Tables S3–S4] Two pieces of the authors' own supplementary data complicate the pODMR narrative and should be confronted in the main text. (i) Table S4 shows that when GPT-5.5 sees all 96 unlabeled measurements in one batch, accuracy is 95.8–100% across all conditions *without* the expected-signal requirement, and the expected-signal condition is not the best at medium reasoning (95.8%, worst row). The main-text claim that requiring an expected-signal calculation is what keeps false positives low therefore appears specific to the zero-shot single-measurement framing; the paper should explain why that framing is the operationally relevant one and reconcile it with the batch result. (ii) Table S3 shows a trivial deterministic contrast-depth threshold (0.132) separates all 96 measurements perfectly. The authors frame this as supporting the division of labor, which is fair, but the stronger implication — t
minor comments (7)
  1. [Abstract / Table 1] The abstract and Introduction describe 'three autonomous experiments,' but one of the three required human advice during reanalysis to reach its final interpretation. A qualifier (e.g., 'two fully autonomous; one completed with human advice during reanalysis') would avoid overstatement at first reading.
  2. [Methods 'Ramsey Checkpoint Benchmark'] Clarify whether the memory snapshot included in each checkpoint package retains the instruction ('do not treat fit success alone as evidence of a resonance') that originally induced the minimum-sampled-point behavior. If so, the benchmark partly measures whether models can overcome a misleading standing instruction — an interesting feature that should be stated explicitly rather than left implicit.
  3. [Fig. 3 caption] State in the caption (not only in Methods) that no confidence intervals are shown and that cells pool 20 replicates; consider annotating the aggregate panel (b) with the per-checkpoint composition caveat, since panel (b) alone invites over-reading.
  4. [Supp. Table S2] Five medium-reasoning domain-facts runs had completed predictions but missing logs (283/288). State explicitly that scoring used the returned judgments and that the missing logs affect only the tool-use audit, not Table S1 counts.
  5. [Throughout] The designation 'GPT-5.6 Sol' is never explained; one sentence identifying the model family and access date would help readers reproduce or contextualize. Similarly, 'mod_depth' appears in the pODMR prompt without definition in the main text.
  6. [Fig. 4] Error bars are described as bootstrap 95% CIs only in Methods; add the resample count (20,000) and the measurement-level resampling scheme to the caption. In panels b–d the y-axis range differs implicitly across conditions in visual salience; a shared scale or explicit note would aid comparison.
  7. [References] Several 2026 preprints (refs. 7, 8, 10, 11, 20) are cited with arXiv numbers embedded in unusual positions and inconsistent formatting; please normalize. Verify that ref. 19 (GitHub) includes a tagged release or commit hash corresponding to this manuscript version for long-term auditability.

Circularity Check

0 steps flagged

No circular derivation: empirical agent demos and offline benchmarks do not reduce predictions to fitted inputs or self-defined quantities.

full rationale

This paper is an experimental systems and evaluation study, not a first-principles derivation of a physical or mathematical result. Its load-bearing claims are (i) that an LLM-plus-deterministic-control workflow can run autonomous NV projects and (ii) that, on two offline benchmarks, reasoning effort and prompt conditions change hypothesis recognition and false-positive rates in measured ways. pODMR labels are assigned from known NV orientation versus whether the targeted transition lies inside the scanned frequency range, independent of model outputs. The Ramsey pass criterion is a fixed external rubric requiring an explicit residual-calibration-offset hypothesis, not a quantity defined by the agent’s own fit parameters. Project records and benchmark packages are released for audit. Constructing checkpoints from one archived run that originally missed the offset is a retrospective design choice that limits generalization, but it does not make reported pass counts equal to their inputs by construction, nor does it rest on a self-citation uniqueness theorem or renamed known law. No self-definitional loop, fitted-input-as-prediction, or load-bearing self-citation chain is present.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 2 invented entities

The central systems claim rests on standard NV measurement physics, a software safety boundary the authors impose, and evaluation design choices (manual hypothesis rubric; orientation-based resonance labels). No new physical entities are postulated. Load-bearing non-derived pieces are the agent–queue architecture assumptions and the benchmark scoring/label conventions rather than fitted universal constants.

free parameters (3)
  • Ramsey pass/fail manual scoring rubric = pre-fixed rubric applied to 1,200 runs
    Binary human judgment requires an explicit residual calibration/microwave-frequency offset hypothesis; nearby language about mismatch or unspecified detuning fails. This threshold defines the benchmark’s main dependent variable.
  • LLM reasoning effort settings (low/medium/high/xhigh) = four discrete levels per model
    Vendor reasoning knobs are experimental factors that drive reported pass-rate and false-positive trends; they are chosen controls, not derived quantities.
  • Retrospective pODMR contrast-depth threshold 0.132 = 0.132 contrast depth
    Supplementary deterministic separator (60% of nominal 22% contrast) cleanly splits this dataset but was not used in the LLM benchmark; still a hand-chosen cutoff on this sample.
axioms (5)
  • domain assumption NV electronic spin can be optically initialized/read out and driven by near-resonant microwave π pulses so pODMR/Ramsey/CPMG encode resonance and coherence information.
    Background quantum-sensing physics assumed throughout case studies and pODMR labeling (Introduction/Methods).
  • domain assumption Deterministic verifier plus shared-folder job queue can enforce parameter bounds and safety so the LLM never needs direct instrument-driver access.
    Architectural premise of the workflow (Fig. 1; Deterministic Hardware Control).
  • domain assumption pODMR measurements from field-aligned vs other-orientation NVs are correctly labeled resonance-present/absent by whether the targeted transition lies inside the 3.825–3.925 GHz scan.
    Ground truth for the classification benchmark (pODMR Data Evaluation Benchmark).
  • ad hoc to paper Persistent markdown project state, memory, and knowledge files plus tool-using LLM calls are sufficient context machinery for multi-hour experimental orchestration.
    Systems assumption specific to this workflow implementation; not independently validated beyond three runs.
  • ad hoc to paper A response passes the Ramsey benchmark only if it links residual Ramsey frequency offset specifically to resonance/microwave calibration error.
    Scoring axiom defining the hypothesis-formation metric (Methods; Table 2).
invented entities (2)
  • Ramsey checkpoint benchmark packages (cp01–cp05) independent evidence
    purpose: Freeze post-Ramsey project state to test whether agents propose residual calibration-offset hypotheses without live lab execution.
    Methodological construct built from one archived autonomous project; useful evaluation device, not a physical object.
  • Agentic NV experiment workflow (LLM orchestration + deterministic hardware gate) independent evidence
    purpose: Separate scientific reasoning from safety-critical instrument control in autonomous NV sensing.
    Systems architecture demonstrated in three case studies; independent evidence is the released logs/data rather than external replication yet.

pith-pipeline@v1.2.0-grok45-kimik3 · 21206 in / 3532 out tokens · 76510 ms · 2026-07-31T00:08:46.649654+00:00 · methodology

0 comments
read the original abstract

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 \(T_2^\ast\) with Ramsey measurements, and added a Carr--Purcell--Meiboom--Gill (CPMG) measurement to check a weak feature that could be related to nearby \(^{13}\mathrm{C}\). 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.

Figures

Figures reproduced from arXiv: 2607.25145 by Mingda Li, Nutdech Phadetsuwannukun, Paola Cappellaro, Ryotaro Okabe, Takuya Isogawa.

Figure 1
Figure 1. Figure 1: Overview of the autonomous NV experiment workflow. The human researcher gives an experimental objective to the LLM agent. The agent uses project records, memory and knowledge documents, and quantitative analysis tools on the agent computer. It submits jobs through a shared folder and receives recorded data from the experiment system. On the experiment computer, deterministic software manages the queue, ver… view at source ↗
Figure 2
Figure 2. Figure 2: Autonomous workflow demonstration from the main experiment. Panel (a) shows the main experimental stages as exchanges between the agent and the experiment system. Panel (b) shows the new image scan. Panels (c) and (d) show strong-π pODMR measurements for the rejected and selected candidates. Panel (e) shows the weak-π pODMR resonance calibration. Panel (f) shows one representative Ramsey measurement, and p… view at source ↗
Figure 3
Figure 3. Figure 3: Ramsey checkpoint benchmark across three models. Panel (a) illustrates hypothesis formation from project records and returned data. Panel (b) shows aggregate pass rates over five checkpoints and twenty replicates per checkpoint for GPT-5.4, GPT-5.5, and GPT-5.6 Sol at each reasoning effort. Labels give passing runs out of 100. Panels (c), (d), and (e) show checkpoint level pass counts for GPT-5.4, GPT-5.5,… view at source ↗
Figure 4
Figure 4. Figure 4: pODMR benchmark across three models. Panel (a) illustrates how calculations and simulations guide pODMR data evaluation, with representative measurements without and with a resonance. Panels (b), (c), and (d) show the false positive rate for the sequence, domain facts, and expected signal conditions, respectively. Panels (e), (f), and (g) show the corresponding overall classification accuracy. Each point p… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

22 extracted references · 3 canonical work pages

  1. [1]

    A., MacKnight, R., Kline, B

    Boiko, D. A., MacKnight, R., Kline, B. & Gomes, G. Autonomous chemical research with large language models.Nature 624, 570–578, DOI: 10.1038/s41586-023-06792-0 (2023)

    doi:10.1038/s41586-023-06792-0 2023

  2. [2]

    Zhang, Z.et al.A multimodal robotic platform for multi-element electrocatalyst discovery.Nature647, 390–396, DOI: 10.1038/s41586-025-09640-5 (2025)

    work page doi:10.1038/s41586-025-09640-5 2025

  3. [3]

    Commun.16, 9104, DOI: 10.1038/s41467-025-64105-7 (2025)

    Mandal, I.et al.Evaluating large language model agents for automation of atomic force microscopy.Nat. Commun.16, 9104, DOI: 10.1038/s41467-025-64105-7 (2025)

    doi:10.1038/s41467-025-64105-7 2025

  4. [4]

    E.et al.A multi-agent system for automating scientific discovery.Nature655, 497–505, DOI: 10.1038/ s41586-026-10652-y (2026)

    Ghareeb, A. E.et al.A multi-agent system for automating scientific discovery.Nature655, 497–505, DOI: 10.1038/ s41586-026-10652-y (2026)

    2026

  5. [5]

    Gottweis, J.et al.Accelerating scientific discovery with co-scientist.Nature655, 487–496, DOI: 10.1038/ s41586-026-10644-y (2026)

    2026

  6. [6]

    7.Li, S.et al.Large language model-assisted superconducting qubit experiments (2026)

    Cao, S.et al.Automating quantum computing laboratory experiments with an agent-based ai framework.Patterns6, DOI: 10.1016/j.patter.2025.101372 (2025). 7.Li, S.et al.Large language model-assisted superconducting qubit experiments (2026). 2603.08801

    arXiv 2025

  7. [8]

    2604.25884

    Cao, S.et al.Qcaleval: Benchmarking vision-language models for quantum calibration plot understanding (2026). 2604.25884. 9.Sha, R.et al.Llm-based multi-agent copilot for quantum sensor (2025). 2508.05421

    Pith/arXiv arXiv 2026

  8. [10]

    Wang, D., Qi, L., Xie, Y ., Linke, N. M. & Brown, K. R. A hardware-safety-gated system for llm-written native artiq control code on a trapped-ion platform (2026). 2606.27231

    Pith/arXiv arXiv 2026

  9. [11]

    A., Bright-Thonney, S., Novak, A., Garcia, D

    Moreno, E. A., Bright-Thonney, S., Novak, A., Garcia, D. & Harris, P. Ai agents can already autonomously perform experimental high energy physics (2026). 2603.20179

    Pith/arXiv arXiv 2026

  10. [12]

    Chen, Z.et al.An agentic artificially intelligent x-ray scientist.Nat. Mach. Intell.DOI: 10.1038/s42256-026-01261-5 (2026)

    work page doi:10.1038/s42256-026-01261-5 2026

  11. [13]

    R.et al.Nanoscale magnetic sensing with an individual electronic spin in diamond.Nature455, 644–647, DOI: 10.1038/nature07279 (2008)

    Maze, J. R.et al.Nanoscale magnetic sensing with an individual electronic spin in diamond.Nature455, 644–647, DOI: 10.1038/nature07279 (2008). 7/11

    doi:10.1038/nature07279 2008

  12. [14]

    F.et al.Sensitivity optimization for nv-diamond magnetometry.Rev

    Barry, J. F.et al.Sensitivity optimization for nv-diamond magnetometry.Rev. Mod. Phys.92, 015004, DOI: 10.1103/ RevModPhys.92.015004 (2020)

    2020

  13. [15]

    L., Reinhard, F

    Degen, C. L., Reinhard, F. & Cappellaro, P. Quantum sensing.Rev. Mod. Phys.89, 035002, DOI: 10.1103/RevModPhys. 89.035002 (2017)

    doi:10.1103/revmodphys 2017

  14. [16]

    & Yacoby, A

    Casola, F., van der Sar, T. & Yacoby, A. Probing condensed matter physics with magnetometry based on nitrogen-vacancy centres in diamond.Nat. Rev. Mater.3, 17088, DOI: 10.1038/natrevmats.2017.88 (2018)

    doi:10.1038/natrevmats.2017.88 2017

  15. [17]

    Aslam, N.et al.Quantum sensors for biomedical applications.Nat. Rev. Phys.5, 157–169, DOI: 10.1038/ s42254-023-00558-3 (2023)

    2023

  16. [18]

    Li, S., Jacques, V ., Maletinsky, P., Degen, C. L. & Du, C. R. Multimodal scanning-probe quantum sensing of quantum materials.Nat. Mater.DOI: 10.1038/s41563-026-02648-w (2026)

    work page doi:10.1038/s41563-026-02648-w 2026

  17. [19]

    GitHub repository, https://github.com/takuyaisogawa/nv-autonomous-experiments (2026)

    NV Autonomous Experiments. GitHub repository, https://github.com/takuyaisogawa/nv-autonomous-experiments (2026)

    2026

  18. [20]

    2606.06749

    Nasri, F.et al.Deterministic access to global viral sequence data enables robust agentic scientific discovery (2026). 2606.06749

    Pith/arXiv arXiv 2026

  19. [21]

    Commun.16, 9377, DOI: 10.1038/s41467-025-64430-x (2025)

    Jin, Q.et al.Agentmd: Empowering language agents for risk prediction with large-scale clinical tool learning.Nat. Commun.16, 9377, DOI: 10.1038/s41467-025-64430-x (2025)

    doi:10.1038/s41467-025-64430-x 2025

  20. [22]

    2502.16599

    Schäfermeier, C.et al.Commercial scanning nitrogen vacancy magnetometer in a closed-cycle cryostat (2025). 2502.16599

    Pith/arXiv arXiv 2025

  21. [23]

    R.et al.Micrometer-scale magnetic imaging of geological samples using a quantum diamond microscope

    Glenn, D. R.et al.Micrometer-scale magnetic imaging of geological samples using a quantum diamond microscope. Geochem. Geophys. Geosystems18, 3254–3267, DOI: https://doi.org/10.1002/2017GC006946 (2017). https://agupubs. onlinelibrary.wiley.com/doi/pdf/10.1002/2017GC006946

    doi:10.1002/2017gc006946 2017

  22. [24]

    One plausible physical explanation is residual detuning of roughly 0.5 MHz relative to the weak-pODMR grid center

    Wang, X.et al.Quantum diamond magnetometry for navigation in gnss denied environments. In Freymueller, J. T. & Sánchez, L. (eds.)Gravity, Positioning and Reference Frames, 87–92 (Springer Nature Switzerland, Cham, 2024). Additional information Competing interests:The authors declare no competing interests. 8/11 Table 1.Summary of three autonomous NV cente...

    2024