Pith. sign in

REVIEW 4 major objections 6 minor 35 references

Jointly fitting photometry with astrometry tightens exoplanet orbit posteriors by up to 33%.

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 · deepseek-v4-flash

2026-08-01 06:51 UTC pith:33U62ZSG

load-bearing objection A plausible, well-scoped simulation shows joint photometry+astrometry tightens orbital posteriors, but the headline 12–33% numbers rest on one noise draw and a self-consistent Lambertian model. the 4 major comments →

arxiv 2607.22754 v1 pith:33U62ZSG submitted 2026-07-23 astro-ph.IM astro-ph.EP

Jointly Modeling Roman Coronagraph Astrometry and Photometry Improves Orbital Parameter Estimates

Farrah Molina , Sarah Blunt , Jason Wang This is my paper
classification astro-ph.IM astro-ph.EP
keywords exoplanet direct imagingRoman Coronagraph Instrumentreflected-light photometryorbital parameter estimationBayesian orbit fittingphase functionLambertian scatteringastrometry
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.

The paper tests whether adding reflected-light photometry to astrometric orbit fits improves the recovered orbital parameters for exoplanets imaged by the Roman Coronagraph Instrument. Simulating realistic partial-orbit observations at signal-to-noise ratios of 3, 5, and 10, the authors find that joint fits always tighten the posterior, with inclination precision improving 33% at SNR=10 versus 12% at SNR=3. The gain comes from the phase dependence of reflected light, which breaks degeneracies that astrometry alone leaves. A sympathetic reader would care because Roman is expected to produce the first reflected-light direct imaging of exoplanets, so this is a practical route to better orbits before the Habitable Worlds Observatory.

Core claim

The central claim is that a joint model of astrometric positions and phase-dependent reflected-light brightness, both generated from the same orbital geometry and fit simultaneously, yields narrower posterior distributions for orbital elements than astrometry alone. The effect is SNR-dependent: at SNR=3 photometry adds 12% improvement in inclination precision, while at SNR=10 it adds 33%. The paper further shows that the improvement appears in eccentricity as well as inclination, and that the added constraint grows as data quality improves.

What carries the argument

The load-bearing mechanism is the Lambertian disk reflection model, which gives the planet's brightness as proportional to A α / R², where A is albedo, R is the star-planet distance, and α is a phase-angle function computed from the geometry of the star, planet, and observer. This photometric model is added to an open-source Bayesian orbit-fitting package, so that each orbit hypothesis predicts both an astrometric position and a brightness; the fit then updates the posterior from both data streams. The phase-dependent term is what carries the extra information.

Load-bearing premise

The photometric model assumes a Lambertian disk (matte, uniformly scattering surface) for reflected light, and the same model is used to generate the mock data and to fit them; if real planets reflect light differently, the size of the improvement could change.

What would settle it

Observe a directly imaged exoplanet with Roman at SNR~10 over a partial orbit and compare the inclination posterior from astrometry alone against the joint fit; if the true phase curve deviates from Lambertian in a way that the model cannot absorb, the recovered inclination would be biased or the precision gain would not materialize. A controlled version: generate mock data with a non-Lambertian phase function (e.g., Rayleigh or cloudy scattering) and see whether the Lambertian-based joint fit improves or degrades accuracy.

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

If this is right

  • For Roman Coronagraph Instrument observations, adding photometry to orbit fits should be standard, yielding noticeably tighter inclinations and eccentricities at high SNR.
  • Higher-SNR imaging is especially valuable: the fractional gain from photometry grows from 12% to 33% as SNR goes from 3 to 10, so investments in achieving high SNR pay off doubly.
  • The same joint-fitting approach can be applied to future Habitable Worlds Observatory data, where reflected-light phase curves will be a primary observable.

Where Pith is reading between the lines

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

  • If real directly imaged planets scatter light non-Lambertianly (clouds, Rayleigh scattering, specular highlights), the actual precision gain may differ; a mismatch between model and truth could bias orbital parameters rather than just broaden posteriors.
  • The 33% figure is for one fiducial system (a=50 AU, e=0.3, i=30°); the gain likely depends on orbital phase coverage and viewing geometry, and future work could map where photometry helps most.
  • Joint fitting may help break the known degeneracy between inclination and other orbital elements in partial orbits, which is why inclination improves most.
  • The method assumes Gaussian photon-noise-limited uncertainties; real coronagraphic data include systematics such as speckle and calibration errors that may dilute the gain.

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 / 6 minor

Summary. This manuscript describes an extension to the open-source orbitize! package for jointly fitting astrometric and reflected-light photometric observations of directly imaged exoplanets, using a Lambertian disk phase function. The authors generate mock Roman Coronagraph Instrument (CGI) astrometric and photometric measurements at SNRs of 3, 5, and 10, then run parallel-tempered MCMC orbit fits with and without the photometric likelihood. They report that including photometry improves posterior precision, with the effect increasing with SNR: 12% improvement in inclination precision at SNR=3 and 33% at SNR=10. The Appendix derives the Lambertian brightness model in the orbitize! coordinate system. The paper is framed as a technical Note for the orbitize! community, with the main deliverable being the new joint-modeling capability and a demonstration of its potential value for Roman observations.

Significance. If the quantitative result holds, the paper provides a useful, forward-looking capability for planning Roman CGI observations and for orbit fitting of reflected-light exoplanets. The open-source implementation in orbitize! and the explicit derivation of the Lambertian likelihood are strengths; the paper is reproducible in principle and directly extends a widely used community tool. However, the central quantitative claim—that joint fitting improves posterior precision by 12–33%—rests on a single noise realization per SNR, an undefined precision metric, and the assumption that the Lambertian model used to generate the data also describes the true phase function. These issues do not invalidate the tool itself, but they currently limit the strength of the conclusions about real Roman data. With additional robustness tests and clarified metrics, the paper could be a valuable contribution to the direct-imaging and orbit-fitting literature.

major comments (4)
  1. [Abstract and Section 3] The central quantitative claim, e.g., the '33% improvement in inclination precision at SNR=10', is not defined. It is unclear whether 'improvement' refers to the width of the 68% credible interval, the standard deviation, the variance, or some other measure. The abstract and Section 3 present these numbers as headline results, so the metric must be specified. I request a precise definition and a statement of how the improvement is computed from the posterior samples.
  2. [Sections 2.1 and 3] Each SNR case appears to use a single random noise realization (one mock dataset per SNR). The reported 12% and 33% values are therefore point estimates from one realization and carry no uncertainty. Precision gains could easily vary by several percent across realizations. The authors should repeat the analysis with multiple mock datasets per SNR (e.g., 10–50 random seeds) and report the distribution of improvement, or at least the median and spread. This is load-bearing for the abstract's quantitative claim.
  3. [Section 2.1 and Appendix Eq. (5)] The mock data are generated using the same Lambertian disk model that is subsequently used in the fit. This is a standard self-consistency test, but it means the reported precision gains are conditional on the Lambertian model being exactly correct. The manuscript itself acknowledges in Section 3.1 that the model is 'simplistic' and will be replaced by more realistic atmosphere models. Given that the abstract states the joint fit 'improves orbital parameter estimates' in general terms, I ask for either (a) a sensitivity test with an alternative non-Lambertian phase function (e.g., a Henyey-Greenstein or empirical phase function) to see whether the precision gain persists, or (b) a clear limitation statement in the abstract and conclusions that the quoted gains are valid only under the Lambertian assumption. Without this, the external validity of the central claim is not established.
  4. [Section 2.2] MCMC convergence is assessed 'by eye' (Section 2.2). While the chain lengths are generous, the posterior precision numbers in Section 3 are the main result, and visual convergence checks alone are not sufficient for a quantitative claim. Please provide convergence diagnostics (e.g., Gelman-Rubin statistics, integrated autocorrelation time, or acceptance rates) or, at minimum, overplot traces for a few representative parameters. This concern is secondary to the single-realization issue but still affects the reliability of the quoted improvements.
minor comments (6)
  1. [Appendix Eq. (5)] The symbol R is reused: Eq. (1) defines R as the instantaneous star–planet distance, while the text below Eq. (5) says R is the planetary radius. This is confusing; please use distinct symbols (e.g., r for separation and R_p for planetary radius) and clarify how the relative brightness amplitude is parameterized in the orbitize! implementation.
  2. [Section 2.1] The parameter 'τ 58849 = 0' is not defined. It appears to be the time of periastron passage in MJD, but as written it is unclear. Please define all orbital parameters in the simulation setup.
  3. [Throughout] The term 'Lambertain' is misspelled; it should be 'Lambertian' (Sections 2.1, 3.1, and the Appendix). Also, the abstract says 'Nancy Roman Grace Space Telescope'; the correct name is 'Nancy Grace Roman Space Telescope'.
  4. [References] In the text, both 'Blunt et al. 2020a' and 'Blunt et al. 2020b' are cited, but the reference list appears to give the same journal article twice (2020a and 2020b). If these are the same paper, consolidate; if they are distinct, provide the appropriate bibliographic entries.
  5. [Figure 1] The bottom six panels in Figure 1 are described as showing posterior constraints with and without photometry, but the caption does not indicate which two parameters are plotted (presumably eccentricity and inclination). Please label the panels or expand the caption so the reader can interpret the figure without referring to the text.
  6. [Appendix Eq. (3) and (4)] The derivation would benefit from a sentence defining the physical angles. In particular, α is called 'the angle subtended by the path between the star, the planet, and the observer'—this is the phase angle—and Eq. (4) is the Lambertian phase function. A brief remark connecting β and α to the standard phase-angle notation would make the Appendix clearer.

Circularity Check

0 steps flagged

No circular derivation: the joint-fit improvement is demonstrated with injection-recovery simulations under a model the paper explicitly labels as simplistic; this limits external validity but does not make the result circular.

full rationale

The paper's central claim (joint astrometry+photometry improves orbital posterior precision) is established by simulation, not by definition. Mock data are generated with orbitize! using the Lambertian disk reflection model derived in the Appendix (Eqs. 1-5), and the same model is used in the fit. This is a standard self-consistency test: it validates the pipeline and quantifies the information content of photometry under the assumed phase function, but it does not reduce a prediction to its inputs. The photometric likelihood is a physical forward model (brightness ∝ Aα/R²), not a fitted parameter renamed as a prediction. The paper explicitly acknowledges the limitation in §3.1, calling the Lambertain disk reflection model 'simplistic' and stating that future work will replace it with more realistic atmosphere models. Self-citations to orbitize! (Blunt et al. 2020a, 2024) are normal software citations and are not load-bearing: the conclusion rests on the simulations run in this paper, not on an unverified prior result. No uniqueness theorem, ansatz-by-citation, or definitional equivalence is present. Concerns about non-Lambertian phase curves, single noise realization, and undefined precision metric are correctness/robustness issues, not circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 0 invented entities

The paper's central claim rests on three main presuppositions: the Lambertian reflection law is an adequate description of real reflected-light phase curves, the photon-noise-only uncertainty prescription is realistic for Roman CGI, and the orbit model in orbitize! is correct. The same-model simulation design makes the result a relative information-gain measure, not an absolute accuracy test. One fitted nuisance parameter (relative brightness amplitude) is introduced by the model.

free parameters (1)
  • Relative brightness amplitude (A/R^2) = not reported
    In the current orbitize! implementation, 'we fit only for relative changes in brightness' (Sec. 3, Appendix Eq. 5). The absolute scaling, which folds in albedo and radius, is a fitted/nuisance parameter in the photometric likelihood though its value is not reported.
axioms (4)
  • domain assumption Lambertian disk reflection model describes the reflected-light phase curve
    Appendix Eqs. 3–5 define α and brightness ∝ Aα/R². The paper explicitly calls this assumption 'simplistic' and plans to replace it with realistic atmosphere models in future work.
  • domain assumption Noise is photon-noise limited: σ_ast = λ/D/SNR, σ_phot = flux/SNR
    Eqs. 1–2 in Sec. 2.1 set the uncertainties. No correlated noise, speckle noise, calibration errors, or systematics are included; the magnitude of the photometric improvement depends on this noise model.
  • standard math Keplerian orbit model and coordinate system from orbitize! are correct
    The orbit-fitting framework is inherited from Blunt et al. (2020a) and used without modification; the photometric term is added on top.
  • ad hoc to paper Mock data generated from the same model used for fitting
    The simulation generates astrometry and photometry from a Lambertian model and then fits with the same model (Sec. 2). This is a self-consistency test; it cannot validate the model, only measure information content under the assumed model.

pith-pipeline@v1.3.0-alltime-deepseek · 3718 in / 12535 out tokens · 127512 ms · 2026-08-01T06:51:29.497056+00:00 · methodology

0 comments
read the original abstract

Launching in 2027, the Nancy Roman Grace Space Telescope (Roman) has the potential to directly image exoplanets in reflected light for the first time. Roman imaging will introduce new constraints on exoplanet orbital parameters, since reflected-light intensity depends on orbital phase. In this Note, we discuss an addition to the open-source Python package orbitize!, which allows users to model exoplanet orbits using joint constraints from astrometry and photometric variations due to orbital phase. To investigate the impact of adding photometric data into the orbital model, we simulated realistic measurements of partial orbits, both including and excluding photometry in our model, and computed orbital posteriors. We found that fitting both astrometry and photometry improves posterior precision relative to fitting astrometry alone. This effect was more pronounced for higher-SNR images; for example, photometric data with SNR=10 yielded 33% improvement in inclination precision when including photometry, while SNR=3 data yielded only 12% improvement.

Figures

Figures reproduced from arXiv: 2607.22754 by Farrah Molina, Jason Wang, Sarah Blunt.

Figure 1
Figure 1. Figure 1: The top four panels show 100 random draws from the orbital posterior for our SNR=10 case, jointly fitting astrometry and photometry. The bottom six panels show the posterior orbital constraints with and without photometric measurements for three different SNRs. The true values for the eccentricity and inclination are bolded and outlined in blue to show the improvements in constraint with an increasing SNR … 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

35 extracted references · 2 canonical work pages

  1. [1]

    Space Telescopes and Instrumentation 2022: Optical, Infrared, and Millimeter Wave , year = 2022, editor =

    Roman coronagraph instrument: engineering overview and status. Space Telescopes and Instrumentation 2022: Optical, Infrared, and Millimeter Wave , year = 2022, editor =. doi:10.1117/12.2630540 , adsurl =

    work page doi:10.1117/12.2630540 2022

  2. [2]

    doi:10.17226/26141 , adsurl =

    Pathways to Discovery in Astronomy and Astrophysics for the 2020s. doi:10.17226/26141 , adsurl =

    doi:10.17226/26141

  3. [3]

    10.1051/0004-6361/202244578

    JWST/MIRI coronagraphic performances as measured on-sky , DOI= "10.1051/0004-6361/202244578", url= "https://doi.org/10.1051/0004-6361/202244578", journal =

    doi:10.1051/0004-6361/202244578

  4. [4]

    2011 , eprint=

    Keplerian Orbits and Dynamics of Exoplanets , author=. 2011 , eprint=

    2011

  5. [7]

    , keywords =

    Dynamic temperature selection for parallel tempering in Markov chain Monte Carlo simulations. , keywords =. doi:10.1093/mnras/stv2422 , archivePrefix =. 1501.05823 , primaryClass =

    Pith/arXiv arXiv

  6. [8]

    , keywords =

    First VLTI/GRAVITY Observations of HIP 65426 b: Evidence for a Low or Moderate Orbital Eccentricity. , keywords =. doi:10.3847/1538-3881/ad06b7 , archivePrefix =. 2310.00148 , primaryClass =

    Pith/arXiv arXiv

  7. [10]

    American Astronomical Society Meeting Abstracts \#244 , year = 2024, series =

    The Habitable Worlds Observatory: Status, Plans, and Opportunities. American Astronomical Society Meeting Abstracts \#244 , year = 2024, series =

    2024

  8. [11]

    Hunter, J. D. , Title =. Computing in Science & Engineering , Volume =

  9. [12]

    Space Telescopes and Instrumentation 2020: Optical, Infrared, and Millimeter Wave , year = 2020, editor =

    The Nancy Grace Roman Space Telescope Coronagraph Instrument (CGI) technology demonstration. Space Telescopes and Instrumentation 2020: Optical, Infrared, and Millimeter Wave , year = 2020, editor =. doi:10.1117/12.2562997 , archivePrefix =. 2103.01980 , primaryClass =

    Pith/arXiv arXiv 2020

  10. [13]

    Computing in Science and Engineering , keywords =

    The NumPy Array: A Structure for Efficient Numerical Computation. Computing in Science and Engineering , keywords =. doi:10.1109/MCSE.2011.37 , archivePrefix =. 1102.1523 , primaryClass =

    Pith/arXiv arXiv 2011

  11. [14]

    Nature Methods , keywords =

    SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods , keywords =

  12. [15]

    2026 , note =

    multiprocessing --- Process-based parallelism , howpublished =. 2026 , note =

    2026

  13. [16]

    arXiv , author =:1307.6212 , journal =

    doi:10.1051/0004-6361/201322068 , eid =. arXiv , author =:1307.6212 , journal =

    Pith/arXiv arXiv

  14. [17]

    , keywords =

    The Astropy Project: Building an Open-science Project and Status of the v2.0 Core Package. , keywords =. doi:10.3847/1538-3881/aabc4f , archiveprefix =. 1801.02634 , primaryclass =

    Pith/arXiv arXiv

  15. [18]

    , keywords =

    The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package. , keywords =. doi:10.3847/1538-4357/ac7c74 , archiveprefix =. 2206.14220 , primaryclass =

    Pith/arXiv arXiv

  16. [19]

    2007", month =

    IPython: A System for Interactive Scientific Computing. Computing in Science and Engineering , year = "2007", month = "Jan", volume =. doi:10.1109/MCSE.2007.53 , adsurl =

    doi:10.1109/mcse.2007.53 2007

  17. [20]

    Positioning and Power in Academic Publishing: Players, Agents and Agendas , editor =

    Jupyter Notebooks -- a publishing format for reproducible computational workflows , author =. Positioning and Power in Academic Publishing: Players, Agents and Agendas , editor =

  18. [21]

    doi:10.5281/zenodo.21003741 , url =

    The pandas development team , title =. doi:10.5281/zenodo.21003741 , url =

    work page doi:10.5281/zenodo.21003741

  19. [22]

    Ralf Gommers and Pauli Virtanen and Matt Haberland and Evgeni Burovski and Tyler Reddy and Warren Weckesser and Andrew Nelson and Travis E. Oliphant and David Cournapeau and Ilhan Polat and alexbrc and Pamphile Roy and Lucas Colley and Pearu Peterson and Josh Wilson and Jake Bowhay and endolith and Nikolay Mayorov and Albert Steppi and Stefan van der Walt...

  20. [23]

    Adam Ginsburg and Brigitta Sipőcz and C. E. Brasseur and Madhura Parikh and jcsegovia and Austen Groener and Henrik Norman and derdon and Manon Marchand and Michael Kelley and Sam Bianco and Thomas Robitaille and P. L. Lim and Eero Vaher and Christoph Deil and Michael Mommert and Jenny V Medina and Ricky Nilsson and Erik Tollerud and Matthieu Baumann and ...

  21. [24]

    , archiveprefix = "arXiv", eprint =

    astroquery: An Astronomical Web-querying Package in Python. , archiveprefix = "arXiv", eprint =. doi:10.3847/1538-3881/aafc33 , adsurl =

    doi:10.3847/1538-3881/aafc33

  22. [25]

    Castillo Hair and Remy Prechelt and Marlon and Kyle Barbary and Johannes Buchner and James Tocknell and James Matthews and NeutralKaon and Ian Czekala , title =

    Dan Foreman-Mackey and Adrian Price-Whelan and Will Vousden and Geoffrey Ryan and Matt Pitkin and Víctor Zabalza and jsheyl and Emily Rice and Michael Smith and Leo Singer and Gregory Ashton and Arfon Smith and Kelle Cruz and Zach Weiner and Wolfgang Kerzendorf and Thomas Vandal and Thomas A Caswell and Zhaozhou Li and Stephan Hoyer and Sebastian M. Casti...

  23. [26]

    The Journal of Open Source Software , year = 2016, month = jun, volume =

    corner.py: Scatterplot matrices in Python. The Journal of Open Source Software , year = 2016, month = jun, volume =. doi:10.21105/joss.00024 , adsurl =

    doi:10.21105/joss.00024 2016

  24. [27]

    Behnel and R

    S. Behnel and R. Bradshaw and C. Citro and L. Dalcin and D. S. Seljebotn and K. Smith , journal =. Cython: The Best of Both Worlds , year =. doi:10.1109/MCSE.2010.118 , issn =

    doi:10.1109/mcse.2010.118 2010

  25. [28]

    Dan Foreman-Mackey and Will Meierjurgen Farr and Anne Archibald and Erik Tollerud and David W. Hogg and Andrew Nelson and Nicholas Kern and Jeremy Sanders and Peter Williams and Dustin Lang and Manodeep Sinha and Bence Béky and joezuntz and Adrian Price-Whelan and Eric Dill and Miguel de Val-Borro and Will Vousden and Ilya and Oriol Abril-Pla and Marco Ta...

  26. [29]

    , keywords =

    emcee: The MCMC Hammer. , keywords =. doi:10.1086/670067 , archiveprefix =. 1202.3665 , primaryclass =

    Pith/arXiv arXiv

  27. [30]

    Armando Solé and jialin and Daniel Hay Guest , title =

    Andrew Collette and Thomas Kluyver and Thomas A Caswell and James Tocknell and Jerome Kieffer and Aleksandar Jelenak and Anthony Scopatz and Darren Dale and Chen and Thomas VINCENT and Matt Einhorn and payno and juliagarriga and Pierlauro Sciarelli and Valentin Valls and Satrajit Ghosh and Ulrik Kofoed Pedersen and Mark Kittisopikul and jakirkham and Mart...

  28. [31]

    2013 , publisher =

    Python and HDF5 , author =. 2013 , publisher =

    2013

  29. [32]

    Sarah Blunt and Jason Wang and Henry Ngo and isabelangelo and Devin Cody and TirthDS and vighneshnagpal and Sofia Covarrubias and Thea McKenna and Tomas Stolker and jorgellop and Robert De Rosa and Mireya Arora and Rodrigo Ferrer-Chávez and Malena Rice and William O Balmer and Lea Hirsch and Adam J. R. W. Smith and Dino Chih-Chun Hsu and eshelDror and Mic...

  30. [33]

    The Journal of Open Source Software , keywords =

    orbitize! v3: Orbit fitting for the High-contrast Imaging Community. The Journal of Open Source Software , keywords =. doi:10.21105/joss.06756 , archiveprefix =. 2409.11573 , primaryclass =

    Pith/arXiv arXiv

  31. [34]

    , keywords =

    orbitize!: A Comprehensive Orbit-fitting Software Package for the High-contrast Imaging Community. , keywords =. doi:10.3847/1538-3881/ab6663 , archiveprefix =. 1910.01756 , primaryclass =

    Pith/arXiv arXiv 1910

  32. [35]

    Hanno Rein and Dan Tamayo and Shangfei Liu and Lukas Winkler and Pete Bartram and Ari Silburt and Garett Brown and Stanley A. Baronett and Tiger Lu and Robert Melikyan and Rosemary Dorsey and Francesco Biscani and Steven Rieder and Dave Spiegel and Luke Wriglesworth and Jose Antonio Castro and katvolk and Thomas Rometsch and Dang Pham and Albert Zhang and...

  33. [36]

    , keywords =

    REBOUND: an open-source multi-purpose N-body code for collisional dynamics. , keywords =. doi:10.1051/0004-6361/201118085 , archiveprefix =. 1110.4876 , primaryclass =

    Pith/arXiv arXiv

  34. [37]

    arXiv e-prints , keywords =

    Streamlining and standardizing software citations with The Software Citation Station. arXiv e-prints , keywords =

  35. [38]

    doi:10.5281/zenodo.17654855 , url =

    Tom Wagg and Floor Broekgaarden and Phil Van-Lane and Kai Wu and Kayhan Gültekin , title =. doi:10.5281/zenodo.17654855 , url =

    doi:10.5281/zenodo.17654855