Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Decoding the proton's gluonic density with lattice QCD-informed machine learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A generative inverse mapper decodes the proton's gluon density from lattice QCD and matches global fits within uncertainties.

desk verdict A narrow but honest VAIM demonstration for gluon PDFs; the four-parameter ansatz means the result is a fit within a family, not a full decoding, but the paper deserves serious review as a proof-of-principle. read the letter →

arxiv 2507.17810 v1 pith:HD67YUB7 submitted 2025-07-23 hep-ph hep-latnucl-th

Brandon Kriesten , Alex NieMiera , William Good , T.J. Hobbs , Huey-Wen Lin This is my paper
classification hep-phhep-latnucl-th
keywords gluonPDFlatticeQCDvariationalautoencoderinversemapperpseudo-PDFmethodIoffe-timedistributionproblemgenerativemachinelearningprotonstructure
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

The paper is trying to establish that generative machine learning can solve the inverse problem of extracting the proton's gluon parton distribution function from lattice QCD data. It trains a variational autoencoder inverse mapper to convert reduced pseudo-Ioffe-time distributions (RpITDs), computed on a lattice ensemble with spacing about 0.09 fm and pion mass about 310 MeV, into the parameters of a normalized gluon PDF. The decoded distribution is then compared with five phenomenological global fits and is consistent with them within uncertainties, especially for momentum fractions $0.2 \lesssim x \lesssim 0.7$ where the lattice signals are strongest. If correct, this opens a route for lattice information to enter global analyses directly through a generative, data-driven framework rather than through hand-picked fit forms.

What carries the argument

The central object is the variational autoencoder inverse mapper (VAIM), a variational autoencoder whose latent space is split into a 35-dimensional observable channel and a 256-dimensional generative channel. The observable channel carries the reduced pseudo-Ioffe-time distributions (RpITDs), double ratios of Wilson-line matrix elements normalized to one at zero Ioffe time. Training data are synthetic: PDF parameters $\alpha,\beta,\gamma,\delta$ are drawn from uniform ranges, the four-parameter ansatz $xg(x)/\langle x\rangle_g = x^\alpha(1-x)^\beta(1+\gamma\sqrt{x}+\delta x)/N$ is formed, and RpITDs are generated through the gluon pseudo-PDF matching relation $\mathcal{M}(\nu,z^2)=\int_0^1 dx\,[xg(x)/\langle x\rangle_g]\,R_{gg}(x\nu,z^2\mu^2)$, with the quark-gluon mixing kernel neglected. After training, lattice RpITDs enter the observable channel and the decoder outputs an ensemble of parameter sets sampled from the latent posterior; that ensemble defines the PDF uncertainty. The network uses four residual blocks with skip connections per encoder and decoder, and its three output heads enforce the physical ranges of $\alpha$, $\beta$, and $1+\gamma+\delta>0$.

What would settle it

Generate synthetic RpITDs from a gluon PDF deliberately outside the four-parameter family, such as one with an additional bump at moderate $x$, feed them to the trained VAIM, and check whether the reconstructed RpITDs and predicted PDF reproduce the input within the reported uncertainties; a significant discrepancy would localize the bias in the assumed parameterization rather than the network.

Watch

Extended reading notes

Core claim

The central discovery claimed is that a trained VAIM, supplied with the 35 lattice RpITDs, produces a normalized gluon PDF $xg(x,\mu=2\,\mathrm{GeV})/\langle x\rangle_g$ whose central value lies within the uncertainty bands of the major phenomenological global fits over most of the $x$-range, with the best constraints at $0.2 \lesssim x \lesssim 0.7$. The claim is supported by a closure test: RpITDs reconstructed from the predicted PDF parameters track the lattice input with an average point-by-point deviation below $1\sigma$ and a maximum deviation within $2\sigma$. The paper further claims that Pearson correlations between the 35 input RpITDs and the $x$-dependent PDF reveal a learned latent representation rather than trivial memorization, and that the uncertainty band widens at small $x$ where the lattice constraints are weak.

Load-bearing premise

The load-bearing assumption, which the paper itself flags as an unquantified systematic, is that the true gluon PDF lies inside the four-parameter family $x^\alpha(1-x)^\beta(1+\gamma\sqrt{x}+\delta x)/N$, because the network only sees training data generated from that form and can never return a PDF outside it.

Editorial extensions

If this is right

  • Lattice RpITDs alone, without any experimental input, can produce a gluon PDF whose shape agrees with global fits in the intermediate-to-large-$x$ window, establishing lattice QCD as an independent constraint on the gluon.
  • The VAIM uncertainty band is comparable to phenomenological fits where the lattice data are informative and expands where they are not, so the method reports where the gluon is actually known.
  • The correlation analysis identifies which combinations of Wilson-line length and hadron momentum carry independent information, which can guide future lattice ensembles toward the most constraining kinematic points.
  • Once epistemic uncertainties such as parameterization choice are quantified, improvements in lattice statistics should translate directly into shrinking $x$-dependent PDF uncertainties.

Reading between the lines

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

  • Beyond the paper, replacing the four-parameter ansatz with a wider nonparametric basis would turn the stated unquantified parameterization systematic into a measurable quantity: how much of the decoded shape comes from the assumed functional form versus the lattice data.
  • Beyond the paper, the same observable-to-parton inverse mapping could be applied wherever a perturbative matching relation exists, including polarized gluon distributions, quark helicity distributions, or fragmentation functions.
  • Beyond the paper, a natural endgame is to feed lattice RpITDs directly into a global fit through a covariance-aware likelihood using the trained VAIM as a fast surrogate, rather than comparing decoded PDFs after the fact.
  • Beyond the paper, the small-$x$ widening of the uncertainty band is a concrete prediction that future high-energy electron-ion collider data, which probe that region, could shrink or challenge.
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. The paper presents a variational autoencoder inverse mapper (VAIM) trained on synthetic reduced pseudo-Ioffe-time distributions (RpITDs) generated from a four-parameter gluon PDF ansatz (Eq. (3)) and the gluon pseudo-PDF matching relation (Eq. (2)). After training, the model is fed lattice RpITDs from the MSULat ensemble (a≈0.09 fm, Mπ≈310 MeV) and produces an ensemble of gluon PDF parameters. The resulting normalized gluon PDF xg(x,µ=2 GeV)/<x>g is compared visually with five global fits (CJ22, CT18, JAM22, MSHT20, NNPDF4.0) and claimed to be consistent within uncertainties, especially for 0.2≲x≲0.7. A closure test shows the predicted PDFs regenerate the lattice RpITDs within 2σ, and a Pearson correlation analysis is used to interpret the latent space.

Significance. If the extraction is unbiased, the paper demonstrates a new generative-AI bridge between lattice QCD and phenomenological PDFs, with potential to incorporate lattice data into global analyses. The method is transparent: the training-data generation is explicit, the architecture is described in detail, and the Monte Carlo ensembling for uncertainty propagation is clearly outlined. The paper is honest about unquantified systematics, explicitly listing functional parameterization choice, jackknife-to-MC conversion, and aleatoric/epistemic separation as future work in Sec. 4.1. However, the significance is currently limited by the absence of a misspecification test for the parameterization family and by the lack of any quantitative goodness-of-fit measure in the comparison with global fits.

major comments (3)
  1. [3.2, Eq. (3), Fig. 2] The central claim of consistency with global fits is conditioned on the four-parameter ansatz of Eq. (3), because both the training data and the decoder output are confined to this family. The closure test in Fig. 2 only checks that the predicted PDFs regenerate the lattice RpITDs through the same forward model Eq. (2) used to create the training set, so it cannot detect whether the true gluon PDF lies outside the family. Since Sec. 4.1 explicitly lists "functional parameterization choice" as an unquantified systematic, the paper should either demonstrate that the global-fit PDFs (e.g., CJ22, NNPDF4.0) can be represented by Eq. (3) within the training ranges, or repeat the extraction with a more flexible basis (e.g., a neural-network PDF) and show that the result is stable. Without such a test, the agreement in Fig. 1 may be an artifact of the prior family rather than evidence that the lattice data have been decoded.
  2. [2, Fig. 1] The claim that the VAIM result is "consistent" with CJ22, CT18, JAM22, MSHT20, and NNPDF4.0 is supported only by visual inspection of Fig. 1. Because the five global fits have different uncertainty definitions, a quantitative measure is needed, such as the chi-squared or average |pull| per x-bin between the VAIM band and each fit over the stated 0.2≲x≲0.7 range. Please provide these numbers, including the treatment of correlated uncertainties, or at minimum a table of pointwise deviations. This is load-bearing because "within uncertainties" is the paper's main conclusion.
  3. [4.1, Eq. (4)] The jackknife-to-Gaussian conversion in Eq. (4) is not standard as written. The jackknife variance of the mean is (N−1)/N times the sum of squared deviations of the jackknife samples; if sigma^2_JK denotes this quantity, multiplying by N−1 overestimates the variance by a factor of order N, whereas if sigma^2_JK denotes the sum of squared deviations, the correct scaling would be (N−1)/N, not (N−1). Please clarify the definition of sigma^2_JK and justify the factor, since the width of the VAIM band directly enters the consistency claim.
minor comments (5)
  1. [2, Fig. 2] The statement that the average deviation is below 1σ and the maximum within 2σ needs a precise definition of σ (lattice statistical uncertainty only?) and the number of points; please provide the numerical values.
  2. [3.2, Eq. (3) and decoder description] The constraint 1+γ+δ>0 does not guarantee positivity of the factor 1+γ√x+δx for all x∈[0,1]; since the PDF must be positive, either enforce a stronger condition or verify positivity of all generated PDFs.
  3. [Abstract] The acronym is typeset as "V AIM" with a space in the abstract and elsewhere; this should be corrected to "VAIM".
  4. [1, Introduction] The claim of "first decoding" should be qualified to "first VAIM-based decoding" given previous ML-based gluon PDF extractions in Refs. [31,32].
  5. [4.1, Uncertainty quantification] The uncertainty band is effectively statistical-only; this should be stated in the caption of Fig. 1 to avoid implying that the band includes known systematic uncertainties.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the VAIM extraction is a constrained inversion under a disclosed parametric ansatz, and the agreement with global fits is an independent external validation.

full rationale

The derivation chain is self-contained and not circular. The VAIM is trained on synthetic RpITDs generated from Eq. (2) using the parametric ansatz Eq. (3), then applied to independent lattice RpITDs from Ref. [24]; the output PDF parameters are not obtained by minimizing a chi-square against the lattice data or against the global fits, so the agreement with CJ22, CT18, JAM22, MSHT20, and NNPDF4.0 is an external, non-forced comparison. The closure test in Fig. 2 checks only that the forward model Eq. (2) reproduces the input RpITDs from the predicted parameters; this is a self-consistency condition, not an independent validation of Eq. (3), and the paper explicitly acknowledges this limitation in Sec. 4.1 by listing 'functional parameterization choice' as an unquantified systematic. The self-citations (e.g., [37] for the parameterization) are not load-bearing: the parameterization is disclosed, comparable to phenomenological forms, and no uniqueness theorem is invoked. The restriction of the prediction to the four-parameter family of Eq. (3) is a model-misspecification risk, not a reduction of the prediction to its inputs by construction; therefore no circular step can be exhibited.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central extraction rests on a set of modeling choices: the pseudo-PDF matching relation, the four-parameter ansatz, the single lattice ensemble, and the Jackknife-to-Gaussian conversion. These are not derived in this paper and are only partially validated; the authors list several as unquantified systematics in Sec. 4.1. No new physical entities are introduced.

free parameters (3)
  • Gluon PDF shape parameters α, β, γ, δ = not reported in text
    The final PDF is determined by these four parameters, which are fitted to the 35 lattice RpITDs through the trained VAIM. The functional form Eq. (3) is assumed a priori, so the central result depends on this fitted set.
  • Training prior ranges for α, β, γ, δ = α∈(-0.9,1), β∈(0,10), γ,δ∈(-5,5)
    Hand-chosen uniform priors over the parameterization; they define the family of allowed PDFs and therefore shape the posterior and the uncertainty band.
  • RpITD range cutoff M∈[-0.5,1]
    Ad hoc constraint applied to training data to enforce physical consistency; it affects which synthetic examples the VAIM sees and thus the learned inverse mapping.
assumptions (5)
  • domain assumption Pseudo-PDF matching relation Eq. (2) with the gluon-gluon kernel Rgg from Ref. [18] accurately relates RpITDs to the gluon PDF at this lattice spacing and kinematics.
    The entire training data and inversion rely on this perturbative matching, including neglect of quark-gluon mixing and O(z^2 Λ^2_QCD) corrections (Sec. 3.1).
  • domain assumption The 4-parameter functional form Eq. (3) spans the true gluon PDF behavior over the x-range of interest.
    The VAIM can only output PDFs of this family; if the true gluon PDF lies outside it, the extracted result is biased. The paper acknowledges parameterization uncertainty is unquantified (Sec. 4.1).
  • domain assumption The single lattice ensemble with a≈0.09 fm and Mπ≈310 MeV is sufficiently close to the physical/continuum limit for comparison with phenomenological PDFs at µ=2 GeV.
    No continuum extrapolation is performed in this work; lattice systematics from the single ensemble are not propagated into the uncertainty band.
  • domain assumption The conversion of Jackknife samples to a Gaussian MC ensemble via Eq. (4) faithfully represents the lattice uncertainties for the VAIM input.
    This prescription (Sec. 4.1) is adopted without validation and is listed by the authors as an unquantified systematic.
  • domain assumption The VAIM architecture and training procedure learn a valid inverse mapping from RpITDs to PDF parameters with the chosen priors.
    Relies on standard variational autoencoder theory and the sufficiency of the chosen architecture, demonstrated only by the closure test.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Decoding the proton's gluonic density with lattice QCD-informed machine learning." pith.science (2026). https://pith.science/paper/HD67YUB7

@misc{pith2026250717810,
  author       = {Pith},
  title        = {Pith review of: Decoding the proton's gluonic density with lattice QCD-informed machine learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HD67YUB7}},
  note         = {Machine review of arXiv:2507.17810}
}
abstract

We present a first machine learning-based decoding of the gluonic structure of the proton from lattice QCD using a variational autoencoder inverse mapper (VAIM). Harnessing the power of generative AI, we predict the parton distribution function (PDF) of the gluon given information on the reduced pseudo-Ioffe-time distributions (RpITDs) as calculated from an ensemble with lattice spacing $a\! \approx\! 0.09$ fm and a pion mass of $M_\pi\! \approx\! 310$ MeV. The resulting gluon PDF is consistent with phenomenological global fits within uncertainties, particularly in the intermediate-to-high-$x$ region where lattice data are most constraining. A subsequent correlation analysis confirms that the VAIM learns a meaningful latent representation, highlighting the potential of generative AI to bridge lattice QCD and phenomenological extractions within a unified analysis framework.

Figures

Figures reproduced from arXiv: 2507.17810 by the authors.

Figure 1
Figure 1. Upper: Comparison of the proton’s gluon PDF, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Closure test comparing lattice RpITDs (data points) to the RpITDs [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Pearson Monte Carlo correlations between the latent RpITDs and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Fant\^omas Unconfined: global QCD fits with B\'ezier parameterizations

    hep-ph 2025-07 conditional novelty 4.0 of 10

    A C++ toolkit, Fantomas, uses Bezier control-point parameterizations to generate and combine diverse PDF fits within xFitter.

Reference graph

Works this paper leans on

52 extracted references · 4 canonical work pages · cited by 1 Pith paper

  1. [1]

    Cerutti, A

    M. Cerutti, A. Accardi, I. P. Fernando, S. Li, J. F. Owens, S. Park, Systematic uncertainties from higher-twist corrections in DIS at large x, Phys. Rev. D 111 (9) (2025) 094013. arXiv:2501.06849, doi: 10.1103/PhysRevD.111.094013

    arXiv 2025

  2. [2]

    T. Hou, J. Gao, T. J. Hobbs, K. Xie, S. Dulat, M. Guzzi, J. Hus- ton, P. Nadolsky, J. Pumplin, C. Schmidt, I. Sitiwaldi, D. Stump, C.-P. Yuan, New CTEQ Global Analysis of Quantum Chromodynamics with High-Precision Data from the LHC, Phys. Rev. D 103 (2021) 014013. arXiv:1912.10053, doi:10.1103/PhysRevD.103.014013

    arXiv 2021

  3. [3]

    Cocuzza, W

    C. Cocuzza, W. Melnitchouk, A. Metz, N. Sato, Polarized antimat- ter in the proton from a global QCD analysis, Phys. Rev. D 106 (3) (2022) L031502. arXiv:2202.03372, doi:10.1103/PhysRevD.106. L031502

    arXiv 2022

  4. [4]

    Bailey, T

    S. Bailey, T. Cridge, L. A. Harland-Lang, A. D. Martin, R. S. Thorne, Parton distributions from LHC, HERA, Tevatron and fixed target data: 5 MSHT20 PDFs, Eur. Phys. J. C 81 (4) (2021) 341. arXiv:2012.04684, doi:10.1140/epjc/s10052-021-09057-0

    arXiv 2021

  5. [5]

    R. D. Ball, et al., The path to proton structure at 1% accuracy, Eur. Phys. J. C 82 (5) (2022) 428. arXiv:2109.02653, doi:10.1140/epjc/ s10052-022-10328-7

    arXiv 2022

  6. [6]

    J. J. Ethier, G. Magni, F. Maltoni, L. Mantani, E. R. Nocera, J. Rojo, E. Slade, E. Vryonidou, C. Zhang, Combined SMEFT interpretation of Higgs, diboson, and top quark data from the LHC, JHEP 11 (2021) 089. arXiv:2105.00006, doi:10.1007/JHEP11(2021)089

    arXiv 2021

  7. [7]

    J. Gao, M. Gao, T. J. Hobbs, D. Liu, X. Shen, Simultaneous CTEQ-TEA extraction of PDFs and SMEFT parameters from jet and tt data, JHEP 05 (2023) 003. arXiv:2211.01094, doi:10.1007/JHEP05(2023)003

    arXiv 2023

  8. [8]

    M. N. Costantini, E. Hammou, Z. Kassabov, M. Madigan, L. Mantani, M. Morales Alvarado, J. M. Moore, M. Ubiali, SIMUnet: an open-source tool for simultaneous global fits of EFT Wilson coe fficients and PDFs, Eur. Phys. J. C 84 (8) (2024) 805. arXiv:2402.03308, doi:10.1140/ epjc/s10052-024-13079-9

    arXiv 2024

Show all 52 references
  1. [9]

    Amoroso, et al., Snowmass 2021 Whitepaper: Proton Structure at the Precision Frontier, Acta Phys

    S. Amoroso, et al., Snowmass 2021 Whitepaper: Proton Structure at the Precision Frontier, Acta Phys. Polon. B 53 (12) (2022) 12–A1. arXiv: 2203.13923, doi:10.5506/APhysPolB.53.12-A1

    2022 arXiv

  2. [10]

    Accardi, et al., Strong interaction physics at the luminosity frontier with 22 GeV electrons at Jefferson Lab, Eur

    A. Accardi, et al., Strong interaction physics at the luminosity frontier with 22 GeV electrons at Jefferson Lab, Eur. Phys. J. A 60 (9) (2024) 173. arXiv:2306.09360, doi:10.1140/epja/s10050-024-01282-x

    2024 arXiv

  3. [11]

    Accardi, et al., Electron Ion Collider: The Next QCD Frontier: Un- derstanding the glue that binds us all, Eur

    A. Accardi, et al., Electron Ion Collider: The Next QCD Frontier: Un- derstanding the glue that binds us all, Eur. Phys. J. A 52 (9) (2016) 268. arXiv:1212.1701, doi:10.1140/epja/i2016-16268-9

    2016 arXiv

  4. [12]

    Kriesten, S

    B. Kriesten, S. Liuti, L. Calero-Diaz, D. Keller, A. Meyer, G. R. Gold- stein, J. Osvaldo Gonzalez-Hernandez, Extraction of generalized parton distribution observables from deeply virtual electron proton scattering ex- periments, Phys. Rev. D 101 (5) (2020) 054021. arXiv:1903....

    2020 arXiv

  5. [13]

    Kriesten, P

    B. Kriesten, P. Velie, E. Yeats, F. Y . Lopez, S. Liuti, Parametrization of quark and gluon generalized parton distributions in a dynamical frame- work, Phys. Rev. D 105 (5) (2022) 056022. arXiv:2101.01826, doi: 10.1103/PhysRevD.105.056022

    2022 arXiv

  6. [14]

    H.-W. Lin, W. Melnitchouk, A. Prokudin, N. Sato, H. Shows, First Monte Carlo Global Analysis of Nucleon Transversity with Lattice QCD Con- straints, Phys. Rev. Lett. 120 (15) (2018) 152502. arXiv:1710.09858, doi:10.1103/PhysRevLett.120.152502

    2018 arXiv

  7. [15]

    Hou, H.-W

    T.-J. Hou, H.-W. Lin, M. Yan, C. P. Yuan, Impact of lattice strangeness asymmetry data in the CTEQ-TEA global analysis, Phys. Rev. D 107 (7) (2023) 076018. arXiv:2211.11064, doi:10.1103/PhysRevD.107. 076018

    2023 arXiv

  8. [16]

    Ablat, S

    A. Ablat, S. Dulat, T.-J. Hou, H.-W. Lin, K. Xie, C. P. Yuan, Impact of lattice gluon dataset on CTEQ-TEA global PDFs (2 2025). arXiv: 2502.10630

    2025 arXiv

  9. [17]

    Ji, Parton Physics on a Euclidean Lattice, Phys

    X. Ji, Parton Physics on a Euclidean Lattice, Phys. Rev. Lett. 110 (2013) 262002. arXiv:1305.1539, doi:10.1103/PhysRevLett. 110.262002

    2013 arXiv

  10. [18]

    Balitsky, W

    I. Balitsky, W. Morris, A. Radyushkin, Gluon pseudo-distributions at short distances, SciPost Phys. Proc. 8 (2022) 161. arXiv:2106.01916, doi:10.21468/SciPostPhysProc.8.161

    2022 arXiv

  11. [19]

    Salas-Chavira, Z

    A. Salas-Chavira, Z. Fan, H.-W. Lin, First glimpse into the kaon gluon parton distribution using lattice QCD, Phys. Rev. D 106 (9) (2022) 094510. arXiv:2112.03124, doi:10.1103/PhysRevD.106.094510

    2022 arXiv

  12. [20]

    NieMiera, W

    A. NieMiera, W. Good, H.-W. Lin, Kaon Gluon Parton Distribution and Momentum Fraction from 2 +1+1 Lattice-QCD with High Statistics (6 2025). arXiv:2506.03002

    2025 arXiv

  13. [21]

    Fan, H.-W

    Z. Fan, H.-W. Lin, Gluon parton distribution of the pion from lattice QCD, Phys. Lett. B 823 (2021) 136778. arXiv:2104.06372, doi:10.1016/ j.physletb.2021.136778

    2021 arXiv

  14. [22]

    W. Good, K. Hasan, A. Chevis, H.-W. Lin, Gluon moment and parton distribution function of the pion from Nf =2+1+1 lattice QCD, Phys. Rev. D 109 (11) (2024) 114509. arXiv:2310.12034, doi:10.1103/ PhysRevD.109.114509

    2024 arXiv

  15. [23]

    Khan, et al., Unpolarized gluon distribution in the nucleon from lattice quantum chromodynamics, Phys

    T. Khan, et al., Unpolarized gluon distribution in the nucleon from lattice quantum chromodynamics, Phys. Rev. D 104 (9) (2021) 094516.arXiv: 2107.08960, doi:10.1103/PhysRevD.104.094516

    2021 arXiv

  16. [24]

    Z. Fan, W. Good, H.-W. Lin, Gluon parton distribution of the nucleon from (2+1+1)-flavor lattice QCD in the physical-continuum limit, Phys. Rev. D 108 (1) (2023) 014508. arXiv:2210.09985, doi:10.1103/ PhysRevD.108.014508

    2023 arXiv

  17. [25]

    W. Good, K. Hasan, H.-W. Lin, Toward the first gluon parton distribution from the LaMET, J. Phys. G 52 (3) (2025) 035105.arXiv:2409.02750, doi:10.1088/1361-6471/ada815

    2025 arXiv

  18. [26]

    W. Good, F. Yao, H.-W. Lin, First Nucleon Gluon PDF from Large Mo- mentum Effective Theory (5 2025). arXiv:2505.13321

    2025

  19. [27]

    Egerer, et al., Toward the determination of the gluon helicity dis- tribution in the nucleon from lattice quantum chromodynamics, Phys

    C. Egerer, et al., Toward the determination of the gluon helicity dis- tribution in the nucleon from lattice quantum chromodynamics, Phys. Rev. D 106 (9) (2022) 094511. arXiv:2207.08733, doi:10.1103/ PhysRevD.106.094511

    2022 arXiv

  20. [28]

    Karpie, K

    J. Karpie, K. Orginos, A. Rothkopf, S. Zafeiropoulos, Reconstructing parton distribution functions from Io ffe time data: from Bayesian meth- ods to Neural Networks, JHEP 04 (2019) 057. arXiv:1901.05408, doi:10.1007/JHEP04(2019)057

    2019 arXiv

  21. [29]

    Del Debbio, T

    L. Del Debbio, T. Giani, J. Karpie, K. Orginos, A. Radyushkin, S. Zafeiropoulos, Neural-network analysis of Parton Distribution Func- tions from Ioffe-time pseudodistributions, JHEP 02 (2021) 138. arXiv: 2010.03996, doi:10.1007/JHEP02(2021)138

    2021 arXiv

  22. [30]

    Dutrieux, J

    H. Dutrieux, J. Karpie, K. Orginos, S. Zafeiropoulos, Simple nonpara- metric reconstruction of parton distributions from limited Fourier infor- mation, Phys. Rev. D 111 (3) (2025) 034515. arXiv:2412.05227, doi:10.1103/PhysRevD.111.034515

    2025 arXiv

  23. [31]

    T. Khan, T. Liu, R. S. Sufian, Gluon helicity in the nucleon from lattice QCD and machine learning, Phys. Rev. D 108 (7) (2023) 074502.arXiv: 2211.15587, doi:10.1103/PhysRevD.108.074502

    2023 arXiv

  24. [32]

    T. A. Chowdhury, T. Izubuchi, M. Kamruzzaman, N. Karthik, T. Khan, T. Liu, A. Paul, J. Schoenleber, R. S. Sufian, Polarized and unpolar- ized gluon PDFs: Generative machine learning applications for lattice QCD matrix elements at short distance and large momentum, Phys. Rev. D...

    2025 arXiv

  25. [33]

    Zhang, C

    R. Zhang, C. Honkala, H.-W. Lin, J.-W. Chen, Pion and kaon distribution amplitudes in the continuum limit, Phys. Rev. D 102 (9) (2020) 094519. arXiv:2005.13955, doi:10.1103/PhysRevD.102.094519

    2020 arXiv

  26. [34]

    Boyda, et al., Applications of Machine Learning to Lattice Quantum Field Theory, in: Snowmass 2021, 2022

    D. Boyda, et al., Applications of Machine Learning to Lattice Quantum Field Theory, in: Snowmass 2021, 2022. arXiv:2202.05838

    2021 arXiv

  27. [35]

    Almaeen, Y

    M. Almaeen, Y . Alanazi, N. Sato, W. Melnitchouk, M. P. Kuchera, Y . Li, Variational Autoencoder Inverse Mapper: An End-to-End Deep Learning Framework for Inverse Problems, Proc. Int. Joint Conf. Neural Netw. (IJCNN) (2021) 1–8 arXiv:https://doi.org/ 10.1109/IJCNN52387.2021.95...

    2021

  28. [36]

    Grigsby, B

    J. Grigsby, B. Kriesten, J. Hoskins, S. Liuti, P. Alonzi, M. Burkardt, Deep learning analysis of deeply virtual exclusive photoproduction, Phys. Rev. D 104 (1) (2021) 016001. arXiv:2012.04801, doi:10.1103/ PhysRevD.104.016001

    2021 arXiv

  29. [37]

    Kriesten, T

    B. Kriesten, T. J. Hobbs, Learning PDFs through interpretable latent representations in Mellin space, Phys. Rev. D 111 (1) (2025) 014028. arXiv:2312.02278, doi:10.1103/PhysRevD.111.014028

    2025 arXiv

  30. [38]

    Almaeen, T

    M. Almaeen, T. Alghamdi, B. Kriesten, D. Adams, Y . Li, H.-W. Lin, S. Liuiti, V AIM-CFF: a variational autoencoder inverse mapper solu- tion to Compton form factor extraction from deeply virtual exclusive reactions, Eur. Phys. J. C 85 (5) (2025) 499. arXiv:2405.05826, doi:10.1...

    2025 arXiv

  31. [39]

    Follana, Q

    E. Follana, Q. Mason, C. Davies, K. Hornbostel, G. P. Lepage, J. Shigemitsu, H. Trottier, K. Wong, Highly improved staggered quarks on the lattice, with applications to charm physics, Phys. Rev. D 75 (2007) 054502. arXiv:hep-lat/0610092, doi:10.1103/PhysRevD. 75.054502

    2007 arXiv

  32. [40]

    Bazavov, et al., Lattice QCD Ensembles with Four Flavors of Highly Improved Staggered Quarks, Phys

    A. Bazavov, et al., Lattice QCD Ensembles with Four Flavors of Highly Improved Staggered Quarks, Phys. Rev. D 87 (5) (2013) 054505.arXiv: 1212.4768, doi:10.1103/PhysRevD.87.054505

    2013 arXiv

  33. [41]

    Delmar, C

    J. Delmar, C. Alexandrou, K. Cichy, M. Constantinou, K. Hadjiyian- nakou, Gluon PDF of the proton using twisted mass fermions, Phys. Rev. D 108 (9) (2023) 094515. arXiv:2310.01389, doi:10.1103/ PhysRevD.108.094515

    2023 arXiv

  34. [42]

    D. P. Kingma, M. Welling, Auto-Encoding Variational Bayes, arXiv preprint (Dec 2013). arXiv:1312.6114

    2013 arXiv

  35. [43]

    Kriesten, J

    B. Kriesten, J. Gomprecht, T. J. Hobbs, Explainable AI classification for parton density theory, JHEP 11 (2024) 007. arXiv:2407.03411, doi: 10.1007/JHEP11(2024)007. 6

    2024 arXiv

  36. [44]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recog- nition, CoRR abs/1512.03385 (2015). arXiv:1512.03385. URL http://arxiv.org/abs/1512.03385

    2015 arXiv

  37. [45]

    Almaeen, J

    M. Almaeen, J. Grigsby, J. Hoskins, B. Kriesten, Y . Li, H.-W. Lin, S. Liuti, Benchmarks for a Global Extraction of Information from Deeply Virtual Exclusive Scattering (7 2022). arXiv:2207.10766

    2022 arXiv

  38. [46]

    L. Kotz, A. Courtoy, P. Nadolsky, F. Olness, M. Ponce-Chavez, Analy- sis of parton distributions in a pion with B ´ezier parametrizations, Phys. Rev. D 109 (7) (2024) 074027. arXiv:2311.08447, doi:10.1103/ PhysRevD.109.074027

    2024 arXiv

  39. [47]

    Kriesten, T

    B. Kriesten, T. J. Hobbs, Anomalous electroweak physics unraveled via evidential deep learning (12 2024). arXiv:2412.16286

    2024 arXiv

  40. [48]

    L. Kotz, A. Courtoy, P. Nadolsky, M. Ponce-Chavez, Fantˆomas: epistemic and nuclear uncertainties for the parton distributions of the pion (5 2025). arXiv:2505.13594

    2025 arXiv

  41. [49]

    Pumplin, D

    J. Pumplin, D. Stump, R. Brock, D. Casey, J. Huston, J. Kalk, H. L. Lai, W. K. Tung, Uncertainties of predictions from parton distribution functions. 2. The Hessian method, Phys. Rev. D 65 (2001) 014013. arXiv:hep-ph/0101032, doi:10.1103/PhysRevD.65.014013

    2001 arXiv

  42. [50]

    Pumplin, D

    J. Pumplin, D. R. Stump, J. Huston, H. L. Lai, P. M. Nadolsky, W. K. Tung, New generation of parton distributions with uncertainties from global QCD analysis, JHEP 07 (2002) 012. arXiv:hep-ph/0201195, doi:10.1088/1126-6708/2002/07/012

    2002 arXiv

  43. [51]

    Hou, et al., Reconstruction of Monte Carlo replicas from Hessian parton distributions, JHEP 03 (2017) 099

    T.-J. Hou, et al., Reconstruction of Monte Carlo replicas from Hessian parton distributions, JHEP 03 (2017) 099. arXiv:1607.06066, doi: 10.1007/JHEP03(2017)099

    2017 arXiv

  44. [52]

    R. G. Edwards, B. Joo, The Chroma software system for lattice QCD, Nucl. Phys. B Proc. Suppl. 140 (2005) 832. arXiv:hep-lat/0409003, doi:10.1016/j.nuclphysbps.2004.11.254. 7

    2005 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.