Pith. sign in

REVIEW 3 major objections 5 minor 57 references

Autonomous discovery of accelerator commissioning algorithms

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A language-model agent, testing its own code in simulation, reduced the mean injections to capture beam from 207.5 to 20.3.

desk verdict A credible proof-of-concept that an LLM loop can improve a simulated commissioning procedure, but the headline factor-of-ten gains are in-sample until a held-out seed check is added. read the letter →

arxiv 2608.07138 v1 pith:WO2EU5Z4 submitted 2026-08-07 physics.acc-ph cs.AI

Thorsten Hellert (Lawrence Berkeley National Laboratory) This is my paper · ORCID
classification physics.acc-phcs.AI
keywords autonomousresearchlooplanguage-modelagentcommissioningalgorithmbeamcapturesimulatedmulti-objectivesearchParetofrontaccelerator
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper demonstrates that a language-model agent can discover and improve accelerator commissioning algorithms on its own, rather than merely execute or tune human-designed procedures. The closed loop proposes a code change, screens it for invalid simulator access, evaluates it on a fixed ensemble of perturbed machines, and merges the change only if it lowers the ensemble-mean cost. Applied to RF beam capture in the ALS-U accumulator-ring model, the best campaign cut the mean number of injections to capture from 207.5 to 20.3, and a stripped-down campaign built a working capture procedure from a minimal stub when given a library of reusable routines. The same loop, run with a Pareto merge rule, produced 16 non-dominated procedures spanning a trade-off between fast capture and correction of seeded machine errors. If this holds, simulated commissioning becomes a mode of autonomous algorithm discovery, not just validation.

What carries the argument

The carrying mechanism is the closed propose–screen–evaluate–merge loop, adapted for accelerator commissioning. A proposer edits the algorithm code in an isolated worktree; an independent reviewer rejects diffs that access unavailable quantities or bypass action costs; the fixed harness evaluates survivors on the same 50-seed ensemble with a cost that penalizes failed seeds by how far they advance; and a deterministic merge predicate keeps a change only if it improves the incumbent score, or, in the multi-objective version, if it is not Pareto-dominated. The key protection is the strict separation between the algorithm under development and the experiment that judges it: the agent may modify the algorithm and its helper library, but not the lattice, seeded errors, simulator state, action costs, capture criterion, or scoring rule.

What would settle it

Take the best retained algorithm from the scalar campaign and run it on a fresh ensemble of 50 error seeds that were never used during search, with the same harness and budget. If the mean injections to capture on those held-out seeds is close to or above the expert baseline of 207.5, the reported improvement is largely an artifact of selecting on the fixed 50-seed ensemble.

Watch

Extended reading notes

Core claim

The central discovery is that treating the commissioning procedure itself as the search object, instead of only tuning machine variables within a fixed procedure, lets an automated loop outperform a published expert baseline and create viable procedures from minimal scaffolding. The evidence is a roughly tenfold reduction in mean injections to capture on a fixed 50-seed ensemble, with the improvement coming mostly from streamlining the expert procedure and one genuinely new recovery move that nudges correctors near the injection point to escape repeated beam loss. The paper also finds a capability threshold among model tiers, a dominant role for the executable helper library in the scaffold ablation, and a Pareto-optimal set of 16 procedures in the two-objective campaign. The author presents this as a proof-of-concept that commissioning studies can be reframed from evaluating human-designed procedures to enabling agents to participate directly in discovering accelerator algorithms.

Load-bearing premise

The load-bearing premise is that the fixed 50-seed error ensemble is representative enough that algorithms selected on it will also perform well on other error realizations, since the same ensemble is used for both the merge predicate and the final scoring.

Editorial extensions

If this is right

  • Simulated commissioning studies can be re-run quickly after lattice or hardware changes, because the loop re-derives procedures instead of requiring expert rewriting.
  • A multi-objective campaign can populate a full trade-off surface of validated procedures in one run, giving operators a menu of options rather than a single hand-built compromise.
  • Commissionability can be assessed earlier in the design cycle, since the loop's monetary cost is low enough for repeated design iteration on a workstation.
  • The loop can be extended to an end-to-end first-injection-to-user-operation procedure, requiring longer evaluations but no fundamental framework change.
  • Agent-discovered procedures can include genuinely new steps, such as the corrector-nudge recovery move, that are not present in the expert baseline.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the same 50 seeds are used for selection and final scoring, I would expect that evaluating retained algorithms on a fresh seed set would reveal some overfitting; the paper does not report such a held-out test.
  • The large effect of the helper library relative to the physics documents suggests that encoding procedural knowledge as executable routines is the most efficient scaffold, and that future autonomous-search systems might benefit from having the agent itself generate and maintain such libraries.
  • An implicit next step is co-design: by letting the loop modify diagnostics, controls, tolerances, or lattice parameters alongside the procedure, one could optimize the accelerator for commissionability, not just the procedure.
  • The Pareto front's endpoints show that the search can surface qualitatively different strategies; a similar loop could be applied to other commissioning stages if scalar rewards can be defined for them.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This Letter reports an empirical study of an "autoresearch" loop for accelerator commissioning procedures. A language-model proposer modifies Python code implementing RF beam capture for the ALS-U accumulator ring, an independent reviewer screens the diff, and a fixed pySC harness evaluates each surviving candidate on a fixed 50-seed ensemble using a predeclared scalar objective: mean injections to capture, with a phase-based partial-credit penalty for failed seeds. A deterministic merge predicate keeps only candidates that improve the incumbent ensemble mean. The main results are that the best Sonnet and Opus campaigns reduce the ensemble-mean score from the expert baseline of 207.5 to 20.3 and 27.5 injections, respectively (Fig. 2); ablations starting from a non-capturing stub show that the helper library is the dominant scaffold and that stronger models compensate more for missing scaffolding (Fig. 3); and a Pareto version yields 16 non-dominated algorithms spanning capture cost versus correction score (Fig. 4). The Letter argues that this reframes simulated commissioning from evaluating human-designed procedures toward autonomously discovering them.

Significance. The demonstration is potentially significant for the accelerator-commissioning community if the results generalize beyond the specific 50-seed ensemble. The paper's strengths are real: the harness-integrity discussion in Appendix C is unusually thorough and candid about reward-hacking failure modes; the evaluation is performed by an external, established simulator (pySC); the merge rule is deterministic; and code, harness, and campaign configurations are openly available with pinned model snapshots. The scaffold ablation is a well-designed 2x2 experiment. However, the practical significance is conditional: because selection and final scoring use the same fixed ensemble, the headline improvements are in-sample estimates, and the manuscript does not yet demonstrate robustness to new error realizations.

major comments (3)
  1. [Sec. II (Merge), Sec. III (Fig. 2)] The same 50-seed ensemble is used both for the merge predicate and for the reported best-so-far curves and final scores. A candidate is accepted only if it improves the ensemble-mean score on that exact ensemble, so the reported reduction from 207.5 to 20.3 injections is an in-sample optimum of a greedy search, not an unbiased estimate for new realizations of the error model. Appendix C thoroughly addresses benchmark-boundary integrity but does not address cross-seed generalization. Please evaluate all final retained algorithms, and ideally the expert baseline, on a fresh held-out seed ensemble drawn from the same error distribution, and report means and seed-level spreads; without this, the practical claim that the loop discovers commissioning algorithms that work on other seeded machines is not yet supported.
  2. [Sec. IV (Fig. 4)] The Pareto front is produced by a single 200-experiment Sonnet campaign, with dominance evaluated on the same 50 seeds used for final reporting. The front's composition, and the claim that a single campaign produces 16 physically meaningful algorithms, therefore has the same in-sample limitation, and there are no replicate campaigns to assess run-to-run variation. Please re-score the 16 retained algorithms on held-out seeds and, ideally, report two or three independent campaigns to show that the front is stable.
  3. [Sec. III (scalar objective) and Appendix A] The scalar objective contains free constants, specifically the 500-injection failed-seed penalty and the 100-injection per-phase partial credit in the formula 500 + 100(6-k), and the ensemble size is fixed at 50, but no sensitivity analysis is reported. Because the merge predicate is defined by the ensemble mean under this objective, the discovered algorithms and the magnitude of the reported improvement could depend on these choices. A small sensitivity study, for example varying the failed-seed penalty and phase weights and perhaps evaluating on a different number of seeds, would establish that the central conclusions are not artifacts of the particular objective weights.
minor comments (5)
  1. [Fig. 3] The row and column labels with plus and minus symbols ("knowledge + helpers +" and similar) are hard to parse; please use explicit labels such as "documents present/absent" and "library present/absent".
  2. [Sec. III, Fig. 2] The headline values 20.3, 27.5, and 53.6 are the best of three campaigns per tier, and the text does not report the full set of campaign outcomes. Since the shaded bands show scatter across campaigns, please also give all per-campaign values or a median with range so readers can assess the model-tier comparison.
  3. [Sec. III, Fig. 3] The scaffold ablation uses only three or four campaigns per condition and reports no statistical test; the qualitative claims about model capability differences should be accompanied by per-campaign values in a table or appendix, or explicitly framed as exploratory.
  4. [Appendix B] The reproduction details pin the model snapshots and SDK version but do not specify the prompt texts or persona definitions, which are likely to affect agent behavior; consider releasing the campaign prompts or a minimal reproducible configuration.
  5. [Sec. III] The expert baseline is a port of the author's own previously published ALS-U procedure [8] and is not optimized for the injection-count objective, as the text acknowledges; stating this caveat in the abstract or introduction would help readers interpret the factor-of-ten improvement correctly.

Circularity Check

0 steps flagged · score 0.0 of 10

No constructed circularity; the baseline and helper library come from the author's prior work, but they are starting points, not load-bearing conclusions, and the improvement is measured by an external simulator.

full rationale

The derivation chain is: port the published ALS-U procedure [8] as baseline and helper library; let a language-model agent propose code modifications; screen the diffs; evaluate on a fixed 50-seed pySC ensemble; merge only if the ensemble-mean cost improves; report the best-so-far cost. The only self-citation is the baseline and helper library being Python ports of the author's own prior paper [8]. That citation supplies the starting point and scaffolding, not the conclusion: the agent's improvements are selected by a deterministic, harness-owned metric against an independent simulator (pySC), and the reported reduction from 207.5 to 20.3 injections is an empirical outcome of the search, not a quantity forced by the merge predicate. The merge rule guarantees monotonic improvement but not its magnitude; the final value depends on the model's proposals. The ablation from a minimal stub includes conditions without the helper library, so the claim of constructing a working procedure from a minimal starting point does not reduce to the cited prior work. The fixed 50-seed ensemble is used for both selection and final scoring, meaning the reported scores are in-sample and could be overfit to those seeds; however, the paper does not present them as predictions on unseen seeds, so this is an external-validity caveat rather than a circular step. Appendix C further separates the agent's objective from the measured quantity by protecting the harness and rejecting reward-hacking attempts. No equation or definition is self-referential, and no load-bearing conclusion is justified solely by a self-citation. Therefore no significant circularity is present.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters listed are benchmark design choices, not fitted quantities. The axioms are domain assumptions about simulator fidelity, error representativeness, success criteria, action pricing, and standard textbook physics.

free parameters (2)
  • Surrogate cost constants (500, 100) = 500 + 100(6-k)
    Used to price failed seeds in the scalar objective; chosen by hand, not fitted. Affects merge decisions for candidates that fail on some seeds, so it can influence which algorithms are retained.
  • Seed ensemble size (50) = 50 fixed seeds
    The fixed set of error realizations. Results are reported on the same ensemble used for search; no held-out set. This is a design choice, not fitted, but it defines the benchmark.
assumptions (5)
  • domain assumption pySC simulator faithfully models ALS-U accumulator-ring beam dynamics
    The entire evaluation rests on the simulator's fidelity. Referenced in Section II and Appendix A.
  • domain assumption The error model (Table I) is representative of real ALS-U commissioning errors
    The seeded errors define the problem. If unrepresentative, the discovered procedures may not transfer to the real machine.
  • domain assumption The capture criterion (80% of 100 particles survive 500 turns) is a valid definition of beam capture
    Used as the harness-owned success criterion; the agent is not allowed to change it.
  • domain assumption The operator interface actions are priced equivalently to real machine time
    The benchmark boundary assumes action costs are physically meaningful. Appendix C describes fixes to enforce this.
  • standard math Standard accelerator physics from the provided textbooks is accurate
    The agent may retrieve from Wiedemann and Di Mitri; the correctness of the domain knowledge is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Autonomous discovery of accelerator commissioning algorithms." pith.science (2026). https://pith.science/paper/WO2EU5Z4

@misc{pith2026260807138,
  author       = {Pith},
  title        = {Pith review of: Autonomous discovery of accelerator commissioning algorithms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WO2EU5Z4}},
  note         = {Machine review of arXiv:2608.07138}
}
read the original abstract

Simulated commissioning has become essential for de-risking modern light-source design and commissioning, but the procedures being simulated are still designed entirely by human experts. Their labor-intensive redevelopment after lattice changes makes such studies hard to repeat and limits their use during early design iteration. This Letter demonstrates a closed research loop in which a language-model agent writes commissioning code, tests it in simulation, and improves the algorithm from the results. Applied to RF beam capture in the ALS-U accumulator-ring model, the loop substantially improves a working expert procedure and can construct a working one from a minimal starting point, with more capable models succeeding from less initial code. Extending the same framework to multiple objectives produces 16 non-dominated algorithms spanning physically distinct trade-offs between rapid beam capture and correction of seeded machine errors. This reframes commissioning studies from evaluating human-designed procedures toward a mode in which agents participate directly in discovering accelerator algorithms.

Figures

Figures reproduced from arXiv: 2608.07138 by the authors.

Figure 1
Figure 1. FIG. 1. The [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Beam-capture capability across model tiers. Solid [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Ablation of domain-knowledge documents and the [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Autonomous exploration of beam-capture cost and [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 47 canonical work pages

  1. [1]

    P. F. Tavares, S. C. Leemann, M. Sj¨ ostr¨ om, and˚A. An- dersson, J. Synchrotron Radiat.21, 862 (2014)

    work page 2014

  2. [2]

    Raimondi, C

    P. Raimondi, C. Benabderrahmane, P. Berkvens, J. C. Biasci, P. Borowiec, J.-F. Bouteille, T. Brochard, N. B. Brookes, N. Carmignani, L. R. Carver, J.-M. Chaize, J. Chavanne, S. Checchia, Y. Chushkin, F. Cianciosi, M. Di Michiel, R. Dimper, A. D’Elia, D. Einfeld, F. Ewald, L. Farvacque, L. Goirand, L. Hardy, J. Jacob, L. Jolly, M. Krisch, G. Le Bec, I. Lec...

    work page 2023

  3. [3]

    Sajaev, M

    V. Sajaev, M. Borland, J. Calvey, J. Dooling, L. Emery, K. Harkay, N. Kuklev, O. Mohsen, Y. Sun, N. Arnold, T. Berenc, A. Brill, H. Bui, J. Carwardine, W. Cheng, T. Fors, M. Kelly, R. Lindberg, G. Shen, M. Smith, F. Rafael, H. Shang, R. Soliday, U. Wienands, and B. Yang, inProc. NAPAC’25(JACoW Publishing, Geneva, Switzerland, 2025) pp. 1–6

    work page 2025

  4. [4]

    Steier, J

    C. Steier, J. Bohon, S. Borra, K. Chow, C. Espino- Devine, E. DiMasi, G. Ganetis, T. Hellert, M. Johans- son, J. Joseph, J.-Y. Jung, R. Leftwich-Vann, D. Leit- ner, H. Lee, A. Lodge, M. Lerche, T. Luo, R. Miller, D. Nett, B. Nicquevert, O. Omolayo, D. Paulic, A. Ratti, D. Robin, C. Sun, C. Swenson, S. Trovati, M. Venturini, W. Waldron, E. Wallen, and D. W...

    work page 2025

  5. [5]

    C. G. Schroer, I. Agapov, W. Brefeld, R. Brinkmann, Y.-C. Chae, H.-C. Chao, M. Eriksson, J. Keil, X. Nuel Gavald` a, R. R¨ ohlsberger, O. H. Seeck, M. Sprung, M. Tischer, R. Wanzenberg, and E. Weck- ert, J. Synchrotron Radiat.25, 1277 (2018)

    work page 2018

  6. [6]

    Borland, G

    M. Borland, G. Decker, L. Emery, V. Sajaev, Y. Sun, and A. Xiao, J. Synchrotron Radiat.21, 912 (2014)

    work page 2014

  7. [7]

    Sajaev, Phys

    V. Sajaev, Phys. Rev. Accel. Beams22, 040102 (2019)

    work page 2019

  8. [8]

    Hellert, P

    T. Hellert, P. Amstutz, C. Steier, and M. Venturini, Phys. Rev. Accel. Beams22, 100702 (2019)

    work page 2019

Show all 57 references
  1. [9]

    Apollonio, R

    M. Apollonio, R. Fielder, H. Ghasem, and I. Martin, in Proc. IPAC’21(JACoW Publishing, Geneva, Switzer- land, 2021) pp. 265–268

    2021

  2. [10]

    Amorim, A

    D. Amorim, A. Loulergue, L. S. Nadolski, and R. Na- gaoka, inProc. IPAC’21(JACoW Publishing, Geneva, Switzerland, 2021) pp. 171–174

    2021

  3. [11]

    Blanco-Garc ´ ıa, D

    O. Blanco-Garc ´ ıa, D. Amorim, M. Deniaud, A. Louler- gue, L. Nadolski, and R. Nagaoka, inProc. IPAC’22(JA- CoW Publishing, Geneva, Switzerland, 2022) pp. 433– 436

    2022

  4. [12]

    Hellert, I

    T. Hellert, I. Agapov, S. Antipov, R. Bartolini, R. Brinkmann, Y.-C. Chae, D. Einfeld, M. Jebram- cik, and J. Keil, inProc. IPAC’22(JACoW Publishing, Geneva, Switzerland, 2022) pp. 1442–1444

    2022

  5. [13]

    Habet, A

    S. Habet, A. Loulergue, L. Nadolski, P. Brunelle, and S. Ducourtieux, inProc. IPAC’25(JACoW Publishing, Geneva, Switzerland, 2025) pp. 698–701

    2025

  6. [14]

    K. Chen, Z. Wang, G. Wang, T. He, Z. Wang, D. He, M. Hosaka, and W. Xu, inProc. IPAC’24(JACoW Pub- lishing, Geneva, Switzerland, 2024) pp. 2043–2045

    2024

  7. [15]

    Benedetti, F

    G. Benedetti, F. Perez, M. Carl` a, O. Blanco-Garc ´ ıa, and Z. Mart ´ ı, inProc. IPAC’25(JACoW Publishing, Geneva, Switzerland, 2025) pp. 2222–2224

    2025

  8. [16]

    X. Wu, L. Tan, S. Xuan, S.-Q. Tian, X. Liu, and Y. Gong, inProc. IPAC’24(JACoW Publishing, Geneva, Switzer- land, 2024) pp. 1342–1345

    2024

  9. [17]

    Chao and I

    H.-C. Chao and I. Martin, inProc. IPAC’24(JACoW Publishing, Geneva, Switzerland, 2024) pp. 1262–1265

    2024

  10. [18]

    Agapov, S

    I. Agapov, S. Antipov, R. Bartolini, R. Brinkmann, Y. Chae, E. C. Cortes-Garcia, D. Einfeld, and T. Hellert, arXiv preprint (2024), arXiv:2408.07995

    2024 arXiv

  11. [19]

    Mart ´ ı, G

    Z. Mart ´ ı, G. Benedetti, M. Carl` a, and U. Iriso, in Proc. IPAC’23(JACoW Publishing, Geneva, Switzer- land, 2023) pp. 3104–3107

    2023

  12. [20]

    D. Ji, B. Wang, and X. Cui, inProc. IPAC’23(JACoW Publishing, Geneva, Switzerland, 2023) pp. 492–495

    2023

  13. [21]

    B. Wang, Z. Duan, D. Ji, Y. Jiao, and Y. Zhao, in Proc. IPAC’21(JACoW Publishing, Geneva, Switzer- land, 2021) pp. 1349–1351

    2021

  14. [22]

    Hellert, C

    T. Hellert, C. Steier, and J. Keil, inProc. IPAC’23 (2023)

    2023

  15. [23]

    Huang and J

    X. Huang and J. Safranek, Phys. Rev. ST Accel. Beams 18, 084001 (2015), arXiv:1502.07799

    2015 arXiv

  16. [24]

    Duris, D

    J. Duris, D. Kennedy, A. Hanuka, J. Shtalenkova, A. Ede- len, P. Baxevanis, A. Egger, T. Cope, M. McIntire, S. Ermon, and D. Ratner, Phys. Rev. Lett.124, 124801 (2020), arXiv:1909.05963

    2020 arXiv

  17. [25]

    Roussel, A

    R. Roussel, A. Hanuka, and A. Edelen, Phys. Rev. Accel. Beams24, 062801 (2021), arXiv:2010.09824

    2021 arXiv

  18. [26]

    C. Xu, T. Boltz, A. Mochihashi, A. S. Garcia, M. Schuh, and A.-S. M¨ uller, Phys. Rev. Accel. Beams26, 034601 (2023), arXiv:2211.09504

    2023 arXiv

  19. [27]

    Kaiser, C

    J. Kaiser, C. Xu, A. Eichler, and A. S. Garcia, Phys. Rev. Accel. Beams27, 054601 (2024), arXiv:2401.05815

    2024 arXiv

  20. [28]

    Kaiser, C

    J. Kaiser, C. Xu, A. Eichler, A. S. Garcia, O. Stein, E. Br¨ undermann, W. Kuropka, H. Dinter, F. Mayet, T. Vinatier, F. Burkart, and H. Schlarb, Sci. Rep.14, 15733 (2024), arXiv:2306.03739

    2024 arXiv

  21. [29]

    Roussel, A

    R. Roussel, A. L. Edelen, T. Boltz, D. Kennedy, Z. Zhang, F. Ji, X. Huang, D. Ratner, A. S. Gar- cia, C. Xu, J. Kaiser, A. F. Pousa, A. Eichler, J. O. L¨ ubsen, N. M. Isenberg, Y. Gao, N. Kuklev, J. Mar- tinez, B. Mustapha, V. Kain, C. Mayes, W. Lin, S. M. Liuzzo, J. S. John, ...

    2024 arXiv

  22. [30]

    Kaiser, A

    J. Kaiser, A. Lauscher, and A. Eichler, Sci. Adv.11, eadr4173 (2025), arXiv:2405.08888

    2025 arXiv

  23. [31]

    Mayet, GAIA: A General AI Assistant for Intelligent Accelerator Operations (2024), arXiv:2405.01359 [cs.CL]

    F. Mayet, GAIA: A General AI Assistant for Intelligent Accelerator Operations (2024), arXiv:2405.01359 [cs.CL]

    2024 arXiv

  24. [32]

    A. Sulc, T. Hellert, R. Kammering, H. Hoschouer, and J. S. John, inMachine Learning and the Physical Sci- ences Workshop, NeurIPS 2024(2024) arXiv:2409.06336 [physics.acc-ph]

    2024 arXiv

  25. [33]

    Hellert, D

    T. Hellert, D. Bertwistle, S. C. Leemann, A. Sulc, and M. Venturini, Phys. Rev. Research8, L012017 (2026), arXiv:2509.17255

    2026 arXiv

  26. [34]

    Hellert, J

    T. Hellert, J. Montenegro, and A. Sulc, Osprey: Production-Ready Agentic AI for Safety-Critical Control Systems (2025), arXiv:2508.15066 [cs.MA]

    2025

  27. [35]

    A. Karpathy,autoresearch: AI agents running research on single-GPU nanochat training automatically,https: //github.com/karpathy/autoresearch(2026), open- source greedy experiment loop for automated ML re- search

    2026

  28. [36]

    C. Lu, C. Lu, R. T. Lange, J. Foerster, J. Clune, and D. Ha, The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery (2024), arXiv:2408.06292 [cs.AI]

    2024 arXiv

  29. [37]

    D. A. Boiko, R. MacKnight, B. Kline, and G. Gomes, Nature624, 570 (2023)

    2023

  30. [38]

    A. M. Bran, S. Cox, O. Schilter, C. Baldassari, A. D. White, and P. Schwaller, Nat. Mach. Intell.6, 525 (2024), arXiv:2304.05376

    2024 arXiv

  31. [39]

    Romera-Paredes, M

    B. Romera-Paredes, M. Barekatain, A. Novikov, M. Ba- log, M. P. Kumar, E. Dupont, F. J. R. Ruiz, J. S. Ellen- berg, P. Wang, O. Fawzi, P. Kohli, and A. Fawzi, Nature 625, 468 (2024)

    2024

  32. [40]

    Y. J. Ma, W. Liang, G. Wang, D.-A. Huang, O. Bastani, D. Jayaraman, Y. Zhu, L. Fan, and A. Anandkumar, in Int. Conf. on Learning Representations (ICLR)(2024) arXiv:2310.12931 [cs.LG]

    2024 arXiv

  33. [41]

    B. P. MacLeod, F. G. L. Parlane, T. D. Morrissey, F. H¨ ase, L. M. Roch, K. E. Dettelbach, R. Moreira, L. P. E. Yunker, M. B. Rooney, J. R. Deeth, V. Lai, G. J. Ng, H. Situ, R. H. Zhang, M. S. Elliott, T. H. Ha- ley, D. J. Dvorak, A. Aspuru-Guzik, J. E. Hein, and C. P. Berling...

    2020

  34. [42]

    N. J. Szymanski, B. Rendy, Y. Fei, R. E. Kumar, T. He, D. Milsted,et al., Nature624, 86 (2023)

    2023

  35. [43]

    Abolhasani and E

    M. Abolhasani and E. Kumacheva, Nat. Synth.2, 483 (2023)

    2023

  36. [44]

    Stach, B

    E. Stach, B. DeCost, A. G. Kusne, J. Hattrick-Simpers, K. A. Brown, K. G. Reyes,et al., Matter4, 2702 (2021)

    2021

  37. [45]

    Steier, P

    C. Steier, P. Amstutz, K. Baptiste, P. Bong, E. Buice, P. Casey, K. Chow, R. Donahue, M. Ehrlichman, J. Harkins, T. Hellert, M. Johnson, J.-Y. Jung, S. Lee- 9 mann, R. Leftwich-Vann, D. Leitner, T. Luo, O. Omo- layo, J. Osborn, G. Penn, G. Portmann, D. Robin, F. Sannibale, S. ...

    2019

  38. [46]

    Hellert,autoresearch: autonomous discovery of ac- celerator commissioning algorithms,https://github

    T. Hellert,autoresearch: autonomous discovery of ac- celerator commissioning algorithms,https://github. com/als-apg/autoresearch-commissioning(2026), re- search loop, simulated-commissioning harness, and cam- paign configurations

    2026

  39. [47]

    Malina, I

    L. Malina, I. Agapov, J. Keil, E. S. H. Musa, B. Veg- lia, N. Carmignani, L. Carver, L. Hoummi, S. M. Li- uzzo, T. Perron, S. White, and T. Hellert, inProc. ICALEPCS’23(JACoW Publishing, Geneva, Switzer- land, 2023) pp. 1637–1642

    2023

  40. [48]

    S. M. Liuzzo, N. Carmignani, L. R. Carver, L. Hoummi, T. Perron, S. White, I. Agapov, M. Boese, T. Hellert, J. Keil, L. Malina, E. Musa, and B. Veglia, J. Phys. Conf. Ser.2687, 032001 (2024)

    2024

  41. [49]

    Di Mitri,Fundamentals of Particle Accelerator Physics, Graduate Texts in Physics (Springer Cham, 2023)

    S. Di Mitri,Fundamentals of Particle Accelerator Physics, Graduate Texts in Physics (Springer Cham, 2023)

    2023

  42. [50]

    Wiedemann,Particle Accelerator Physics, 4th ed., Graduate Texts in Physics (Springer Cham, 2015)

    H. Wiedemann,Particle Accelerator Physics, 4th ed., Graduate Texts in Physics (Springer Cham, 2015)

    2015

  43. [51]

    Anthropic, System card: Claude Haiku 4.5, https://assets.anthropic.com/m/99128ddd009bdcb/ original/Claude-Haiku-4-5-System-Card.pdf(2025)

    2025

  44. [52]

    Anthropic, System card: Claude Sonnet 4.6,https: //www.anthropic.com/research/claude-sonnet-4-6 (2026)

    2026

  45. [53]

    anthropic.com/news/claude-opus-4-6(2026)

    Anthropic, System card: Claude Opus 4.6,https://www. anthropic.com/news/claude-opus-4-6(2026)

    2026

  46. [54]

    Deb,Multi-Objective Optimization Using Evolution- ary Algorithms(John Wiley & Sons, Chichester, 2001)

    K. Deb,Multi-Objective Optimization Using Evolution- ary Algorithms(John Wiley & Sons, Chichester, 2001)

    2001

  47. [55]

    Amodei, C

    D. Amodei, C. Olah, J. Steinhardt, P. Chris- tiano, J. Schulman, and D. Man´ e, arXiv preprint arXiv:1606.06565 (2016), arXiv:1606.06565 [cs.AI]

    2016 arXiv

  48. [56]

    Leike, M

    J. Leike, M. Martic, V. Krakovna, P. A. Ortega, T. Everitt, A. Lefrancq, L. Orseau, and S. Legg, arXiv preprint arXiv:1711.09883 (2017), arXiv:1711.09883 [cs.LG]

    2017 arXiv

  49. [57]

    Skalse, N

    J. Skalse, N. H. R. Howe, D. Krasheninnikov, and D. Krueger, Advances in Neural Information Processing Systems35, 9460 (2022)

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.