Pith. sign in

REVIEW 5 major objections 5 minor 80 references

Toward an event-level analysis of hadron structure using differential programming

T0 review · 5 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 This paper introduces a differentiable sampling algorithm, LOITS, that lets gradient-based optimization flow from collision events back to the parameters of hadron-structure distributions, and validates it by training a GAN to recover a…

desk verdict LOITS is a plausible but oversold method for differentiable sampling; the gradient derivation is flawed and the closure test doesn't validate the sampling distribution. read the letter →

arxiv 2507.15768 v1 pith:5633IWZ7 submitted 2025-07-21 hep-ph hep-ex

Kevin Braga , Markus Diefenthaler , Steven Goldenberg , Daniel Lersch , Yaohang Li , Jian-Wei Qiu , Kishansingh Rajput , Felix Ringer
show 2 more authors
Nobuo Sato Malachi Schram
This is my paper
classification hep-phhep-ex
keywords differentialprogramminginversetransformsamplingLOITSquantumcorrelationfunctionsgeneralizedpartondistributionsgenerativeadversarialnetworksevent-levelinferencesimulation-based
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

Reconstructing the quark and gluon structure of hadrons from collision data is an inverse problem: the data are individual events, while the quantities of interest are quantum correlation functions (QCFs) such as parton distribution functions, transverse-momentum distributions, and generalized parton distributions. The paper's central claim is that this inference can be performed directly on unbinned events by making the sampling step differentiable. Standard Monte Carlo sampling breaks the gradient chain from the model parameters to the loss, while the introduced LOITS algorithm builds a differentiable path from phase-space samples back to the parameters. A closure test with a generative adversarial network recovers a known two-dimensional density from 10,000 to 1,000,000 sampled events, with reconstruction quality improving as data grow. If the claim holds, QCF parameters could be optimized end-to-end from raw event-level data without binned summary statistics or surrogate models.

What carries the argument

The central object is the LOITS sampler. For a density $p(x,y|\theta)$ it fixes orthogonal reference slices $(x_0,y_0)$, treats $p(x,y_0|\theta)\,p(x_0,y|\theta)$ as the local sampling density, builds one-dimensional CDFs along $x$ and $y$ in each segment, and inverts them by local linear interpolation so that a uniform random input $u$ maps to a sample whose position depends analytically on $\theta$. A subsequent accept-reject correction step can restore asymptotic exactness when the product approximation is inaccurate. The work this machinery does is turning a stochastic sampling operation into a differentiable layer that can sit between a neural-network density image and a loss function defined on samples.

What would settle it

Run the closure test on a two-dimensional Gaussian with correlation coefficient near 0.9 and known parameter dependence, then compare the histogram of LOITS samples against the exact density and compare gradients computed through LOITS against finite-difference gradients for a small parameter step; visible histogram bias or gradient disagreement beyond numerical precision would refute the claims of accurate sampling and exact differentiability.

Watch

Extended reading notes

Core claim

The paper proposes that the non-differentiability of sampling is the missing link in event-level QCF inference, and LOITS is designed to remove it. LOITS partitions phase space into orthogonal segments, approximates the target density inside each segment as a product of one-dimensional densities, computes local cumulative distribution functions, and generates samples by inverting those local CDFs with a differentiable interpolation. Because the CDF values depend on the model parameters, the generated samples inherit computable gradients through the chain rule, and the stated goal is to enable exact gradients of the loss with respect to the model parameters. The authors demonstrate in a closure test that a GAN trained through this differentiable sampler reconstructs a beta-like test density with resolution that improves with event count, and they argue the algorithm generalizes to arbitrary dimensions and to other generative or likelihood-based inference settings.

Load-bearing premise

The method stands on the assumption that inside each small phase-space patch the distribution factorizes into separate x and y pieces; when x and y are strongly correlated this product approximation can be badly wrong, and the paper gives no quantitative bound on the resulting bias.

Editorial extensions

If this is right

  • QCF model parameters (PDFs, TMDs, GPDs) could be optimized directly from unbinned event-level data by gradient descent, skipping the traditional binned cross-section extraction and unfolding steps.
  • The differentiable sampler can be inserted into any simulation-based inference pipeline in which the theoretical density depends smoothly on the model parameters, including likelihood-based fits and generative models.
  • Neural-network representations of multidimensional hadron structure, such as pixelated GPD images, can be trained without an explicit likelihood function.
  • The resolution of the reconstructed density grows with available event statistics, giving a concrete handle on how much data a planned experiment needs to image hadron structure at a given scale.

Reading between the lines

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

  • When LOITS is used without its accept-reject correction, the generator is optimized against an approximate product density; practitioners should treat the uncorrected sampler as a biased surrogate and calibrate the bias before trusting the gradients.
  • Strongly correlated, high-dimensional phase spaces will likely require finer segmentation than the smooth closure-test density, so an adaptive segmentation scheme or a dimension-reduction step would be a natural stress test of the method.
  • Combining LOITS with a differentiable detector surrogate would complete the end-to-end pipeline the paper sketches, allowing gradients to flow through detector effects as well as through the phase-space sampling.
  • The product-density construction suggests LOITS could also serve as a proposal generator for other exact Markov-chain samplers, trading some sampling efficiency for reliable unbiasedness.
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

5 major / 5 minor

Summary. The manuscript introduces LOITS (local orthogonal inverse transform sampling), a differentiable sampling algorithm intended to enable gradient-based optimization of quantum correlation functions (QCFs) directly from event-level data. The method segments phase space and approximates a multi-dimensional density as a product of one-dimensional densities, then applies inverse transform sampling with local linear interpolation so that samples inherit analytic dependence on model parameters. The authors present a GAN-based closure test in which the generator produces a pixelated density image, LOITS converts that image into samples, and a discriminator compares those samples with training data drawn from a known two-dimensional density. The paper claims that LOITS computes exact gradients of the loss with respect to the parameters and that the closure test demonstrates accurate reconstruction of the test density.

Significance. The problem addressed is important: end-to-end, event-level inference of QCFs without surrogate models is a genuine gap in the field, and a differentiable sampling step would be a valuable building block for simulation-based inference. The LOITS idea, combining local orthogonal segmentation with inverse transform sampling and a possible Metropolis-Hastings correction, is interesting and potentially useful. The paper also explicitly connects the method to GPD/TMD imaging and the resolution question, which is a strength. However, the manuscript does not currently establish the central claims: the analytic gradient derivation contains mathematical errors, the sampling approximation is not quantitatively controlled, and the validation is only visual and in-sample. As a result, the contribution is at the stage of a promising method with a toy demonstration rather than a demonstrated exact-gradient, unbiased inference tool.

major comments (5)
  1. [Section III, Eq. (4)] The chain rule written in Eq. (4) is not valid: CDF(x,θ) is already the integral of p(z|θ), so ∂CDF/∂θ_i = ∫_0^x (∂p(z|θ)/∂θ_i) dz; there is no functional derivative ∂CDF/∂p(z|θ) appearing in the integrand. Consequently the presented derivation of the exact gradient is incorrect as written. The paper should either state the correct expression or explain that LOITS relies on automatic differentiation through the interpolation in Eq. (7), which would compute gradients of the approximate sampler rather than of the exact inverse CDF.
  2. [Section III, Eq. (6)] Differentiating the identity CDF(xθ(u,θ),θ)=u with respect to θ_i gives ∂xθ/∂θ_i = −(∂CDF/∂θ_i)/(∂CDF/∂x), with the derivatives evaluated at x=xθ(u,θ). The expression in Eq. (6) omits the division by ∂CDF/∂x and the minus sign, and it does not indicate the evaluation point. This is a load-bearing error because the claim that LOITS computes exact gradients rests on this formula; the authors should provide a correct derivation or empirically verify gradients (e.g., against finite differences).
  3. [Section III, local separability approximation] The local separability approximation p(x,y|θ)≈p(x,y0|θ)p(x0,y|θ) is introduced without normalization or a quantitative accuracy bound; for a general correlated density the approximation can be arbitrarily poor. The GAN closure test in Section IV uses LOITS without the MH correction of Eq. (14), so the generator is optimized against samples from the approximate product density. The discriminator therefore constrains the composition LOITS∘G, not the generator image pG itself; a biased sampler can be compensated by a distorted image. The comparison of the generator image to the ground truth in Figs. 5-6 does not establish that event samples are drawn from the inferred density.
  4. [Section IV, closure test evaluation] The closure test is evaluated only by visual comparison and pixel-ratio maps (Figs. 5-6). There is no quantitative metric (e.g., two-sample test, KL divergence, χ² per bin), no uncertainty estimate, and no baseline comparison (for instance, training the same GAN with a non-differentiable but unbiased sampler, or comparing to unbinned maximum likelihood on the parametric form). Without these, the claim that LOITS yields accurate reconstruction and informative gradients is not demonstrated.
  5. [Section III, LOITS+MH differentiability] The claim that LOITS+MH provides a differentiable sampler is not justified. The acceptance step in Eq. (14) involves a discrete accept/reject decision; standard Metropolis–Hastings does not define a pathwise-differentiable map from θ to the chain state, because rejected proposals cause the state to be carried over with zero derivative through the random accept/reject step. The paper should provide a rigorous argument or a reparameterized MH formulation if this claim is retained.
minor comments (5)
  1. [Section III, text near Eq. (3)] The statement 'CDF(0,θ) = CDF(1,θ) = 1' should read CDF(0,θ)=0 and CDF(1,θ)=1; otherwise the boundary values contradict the definition in Eq. (3).
  2. [Section V, Conclusions] The algorithm is called 'Local Orthogonal' in Section III but 'Longitudinal Orthogonal' in Section V; please make the name consistent throughout the paper.
  3. [Figure 3] Figure 3 would benefit from colorbars and normalized histograms so that the sampling fidelity can be judged quantitatively rather than by eye.
  4. [Section IV, GAN setup] The procedure of averaging the generator output over 1,000 latent noise samples before passing it to LOITS is unusual and should be justified; as written, it suggests the generator is trained on an ensemble-averaged image rather than on individual draws, which may affect mode coverage.
  5. [Section IV, training details] Reference [95] is cited for the distributed training strategy, but the connection to the asynchronous generative inverse problem solver is not obvious; please clarify what is taken from that work.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LOITS is an independently defined differentiable-sampling algorithm, and the GAN closure test uses a fixed ground-truth density; overlapping-author citations are not load-bearing.

full rationale

The paper's central object, the LOITS sampler, is constructed in Section III from first principles: inverse transform sampling, local CDF interpolation, and the explicit local-product approximation p(x,y|theta) ≈ p(x,y0|theta)p(x0,y|theta). It then validates this construction in Section IV against the fixed analytic ground truth p(x,y|phi)=N x^{phi0}(1-x)^{phi1} y^{phi2}(1-y)^{phi3}(1+phi4 xy) (Eq. 15), generating training samples from that density and comparing the GAN-reconstructed image to the same analytic density. No model parameter is fitted to a subset of data and then relabeled as a prediction; the ground-truth image is not used during training, only samples from it. The approximate nature of LOITS is openly acknowledged in Section III ('the phase space sampling is only approximate'), and the MH accept-reject correction (Eq. 14) is explicitly defined using the true target density, so the unbiased variant is not smuggled in by citation. The only overlapping-author references that appear near the method (Refs. [93] and [95]) are used for a training-distribution strategy and a distributed-training implementation, not as the justification for the algorithm's correctness; neither carries the load of the central claim. There is therefore no reduction of the derivation to its inputs: the approximate sampler, the MH correction, and the GAN objective are distinct objects, and the closure test is an external (in-sample) consistency check with a known target. (A separate concern—the analytic-gradient formulas in Eqs. (4) and (6) are mathematically incorrect—is a correctness/validity risk, not a circularity: the wrong formula is not an input that the paper re-exports as a prediction.)

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central method rests on two numerical approximations, namely the local product decomposition and linear CDF interpolation, plus one standard MCMC correctness condition. None of these is quantitatively validated in the GAN closure test. The physics motivation relies on QCD factorization, but the numerical experiment does not use QCD.

free parameters (1)
  • LOITS segmentation grid resolution = not specified for the GAN test (2D examples show 4x4 and 50x50)
    The accuracy and gradient quality of LOITS depend on the number of local segments K, which is chosen by hand; no adaptive scheme or sensitivity study is provided.
assumptions (5)
  • domain assumption The phase space density is strictly positive on the domain, so the CDF is strictly monotonic and invertible.
    Invoked in Section III to justify a unique inverse CDF. This fails for densities with hard zeros or phase space cuts, which are common in physical cross sections.
  • ad hoc to paper The target density is approximately separable in local orthogonal coordinates: p(x,y) ~ p(x,y0) p(x0,y).
    This is the defining approximation of LOITS, introduced in Section III. It is not derived from physics and can be arbitrarily bad for correlated densities.
  • ad hoc to paper Local linear interpolation of the CDF is accurate enough for the resulting gradients to drive optimization.
    The paper uses linear interpolation for the inverse CDF but gives no error bound or sensitivity analysis for the gradient quality.
  • standard math The Metropolis-Hastings step, when used, converges to the target density because the LOITS proposal density is known.
    This is the standard MH theorem, used in Section III. It is valid assuming p* is exactly the proposal distribution, which the paper does not verify numerically.
  • domain assumption QCD factorization formulas such as Eq. (1) connect QCFs to measured cross sections.
    This motivates the method but is not used in the numerical closure test; the test uses a toy two-dimensional density.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Toward an event-level analysis of hadron structure using differential programming." pith.science (2026). https://pith.science/paper/5633IWZ7

@misc{pith2026250715768,
  author       = {Pith},
  title        = {Pith review of: Toward an event-level analysis of hadron structure using differential programming},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5633IWZ7}},
  note         = {Machine review of arXiv:2507.15768}
}
read the original abstract

Reconstructing the internal properties of hadrons in terms of fundamental quark and gluon degrees of freedom is a central goal in nuclear and particle physics. This effort lies at the core of major experimental programs, such as the Jefferson Lab 12 GeV program and the upcoming Electron-Ion Collider. A primary challenge is the inherent inverse problem: converting large-scale observational data from collision events into the fundamental quantum correlation functions (QCFs) that characterize the microscopic structure of hadronic systems within the theory of QCD. Recent advances in scientific computing and machine learning have opened new avenues for addressing this challenge using deep learning techniques. A particularly promising direction is the integration of theoretical calculations and experimental simulations into a unified framework capable of reconstructing QCFs directly from event-level information. In this work, we introduce a differential sampling method called the local orthogonal inverse transform sampling (LOITS) algorithm. We validate its performance through a closure test, demonstrating the accurate reconstruction of a test distribution from sampled events using Generative Adversarial Networks. The LOITS algorithm provides a central building block for addressing inverse problems involving QCFs and enables end-to-end inference pipelines within the framework of differential programming.

Figures

Figures reproduced from arXiv: 2507.15768 by the authors.

Figure 1
Figure 1. FIG. 1. Schematic illustration of the simulation-based infer [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Illustration of the 2D phase space segmentation for [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Sampling of the half moon distribution. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Schematic representation of the GAN setup using the LOITS algorithm. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. GAN-based image inference of the underlying density supervised on 10k phase space samples. The top row shows the [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Pixel-wise ratios between GAN-generated and ground truth densities from different training runs, using datasets of [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

80 extracted references · 22 canonical work pages

  1. [1]

    , K}.(8) 6 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(4×4) 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(4×4) + MH 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(50×50) FIG

    Divide the domain 0< x <1 intoKcontiguous segments: {xk, xk+1;k= 1, . . . , K}.(8) 6 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(4×4) 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(4×4) + MH 0.2 0.4 0.6 0.8 x 0.2 0.4 0.6 0.8 y LOITS(50×50) FIG. 3. Sampling of the half moon distribution.Left:Reconstructed distribution using the LOITS algorithm with a 4×4 phase sp...

    work page

  2. [2]

    Compute segment probabilities and local densities: mk = Z xk+1 xk dx p(x|θ),(9) pk(x|θ) = 1 mk p(x|θ) Θ(x k < x < xk+1).(10)

    work page

  3. [3]

    , Lfor all theksegments, i.e

    Compute the local CDFs for a subgridx k < xl < xk+1 withl= 1, . . . , Lfor all theksegments, i.e. CDFk(xi, θ) = Z xl xk dx pk(x|θ).(11)

    work page

  4. [4]

    To generate a total ofNsamples, estimate the num- ber of samples that need to be generated for each segment via nk = int(mk ·N),(12) where int(·) denotes the integer rounding opera- tion

    work page

  5. [5]

    xθ = LI(u;X k, Yk(θ)).(13) Here,X k, Yk(θ) represent the subgrid and corre- sponding local CDF values

    For each segment, drawn k uniform samplesu∈ [0,1] and map them to phase space valuesx θ using a local interpolation of the CDF, i.e. xθ = LI(u;X k, Yk(θ)).(13) Here,X k, Yk(θ) represent the subgrid and corre- sponding local CDF values

    work page

  6. [6]

    half-moon

    Aggregate the generated samples from all theK segments. By construction, the collection of samples from all seg- ments is differentiable with respect toθ. Similar to the case without local sampling, the samples at the bound- aries of each segment have vanishing gradients, since their corresponding local CDF values are independent ofθ. We revisit this aspe...

    work page

  7. [7]

    A multidimensional unfolding method based on bayes’ theorem,

    G. D’Agostini, “A multidimensional unfolding method based on bayes’ theorem,”Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment362 no. 2, (1995) 487–498

    work page 1995

  8. [8]

    Factorization of Hard Processes in QCD,

    J. C. Collins, D. E. Soper, and G. F. Sterman, “Factorization of Hard Processes in QCD,”Adv. Ser. Direct. High Energy Phys.5(1989) 1–91, arXiv:hep-ph/0409313

    arXiv 1989

Show all 80 references
  1. [9]

    First Monte Carlo Global Analysis of Nucleon Transversity with Lattice QCD Constraints,

    H.-W. Lin, W. Melnitchouk, A. Prokudin, N. Sato, and H. Shows, “First Monte Carlo Global Analysis of Nucleon Transversity with Lattice QCD Constraints,” Phys. Rev. Lett.120no. 15, (2018) 152502, arXiv:1710.09858 [hep-ph]

    2018 arXiv

  2. [10]

    Confronting lattice parton distributions with global QCD analysis,

    J. Bringewatt, N. Sato, W. Melnitchouk, J.-W. Qiu, F. Steffens, and M. Constantinou, “Confronting lattice parton distributions with global QCD analysis,”Phys. Rev. D103no. 1, (2021) 016003,arXiv:2010.00548 [hep-ph]. [5]Jefferson Lab Angular Momentum (JAM), HadStrucCollaboratio...

    2021 arXiv

  3. [11]

    New results in the CTEQ-TEA global analysis of parton distributions in the nucleon,

    A. Ablatet al., “New results in the CTEQ-TEA global analysis of parton distributions in the nucleon,”Eur. Phys. J. Plus139no. 12, (2024) 1063, arXiv:2408.04020 [hep-ph]. [11]NNPDFCollaboration, R. D. Ballet al., “Determination of the theory uncertainties from missing higher or...

    2024 arXiv

  4. [12]

    First global QCD analysis of the TMD g1T from semi-inclusive DIS data,

    S. Bhattacharya, Z.-B. Kang, A. Metz, G. Penn, and D. Pitonyak, “First global QCD analysis of the TMD g1T from semi-inclusive DIS data,”Phys. Rev. D105 no. 3, (2022) 034007,arXiv:2110.10253 [hep-ph]

    2022 arXiv

  5. [13]

    Combination of aN 3LO PDFs and implications for Higgs production cross-sections at the 13 LHC,

    T. Cridgeet al., “Combination of aN 3LO PDFs and implications for Higgs production cross-sections at the 13 LHC,”J. Phys. G52(2025) 6,arXiv:2411.05373 [hep-ph]. [14]Jefferson Lab Angular Momentum (JAM) Collaboration, P. C. Barry, C.-R. Ji, N. Sato, and W. Melnitchouk, “Global ...

    2025 arXiv

  6. [16]

    Global QCD analysis of spin PDFs in the proton with high-xand lattice constraints,

    C. Cocuzza, N. T. Hunt-Smith, W. Melnitchouk, N. Sato, and A. W. Thomas, “Global QCD analysis of spin PDFs in the proton with high-xand lattice constraints,”arXiv:2506.13616 [hep-ph]

  7. [17]

    Next-to-Next-to-Leading Order Global Analysis of Polarized Parton Distribution Functions,

    I. Borsa, M. Stratmann, W. Vogelsang, D. de Florian, and R. Sassot, “Next-to-Next-to-Leading Order Global Analysis of Polarized Parton Distribution Functions,” Phys. Rev. Lett.133no. 15, (2024) 151901, arXiv:2407.11635 [hep-ph]

    2024 arXiv

  8. [18]

    Pion fragmentation functions at high energy colliders,

    I. Borsa, D. de Florian, R. Sassot, and M. Stratmann, “Pion fragmentation functions at high energy colliders,” Phys. Rev. D105no. 3, (2022) L031502, arXiv:2110.14015 [hep-ph]

    2022 arXiv

  9. [19]

    Bayesian Inferring Nucleon’s Gravitation Form Factors via Near-threshold J/ψPhotoproduction,

    Y. Guo, F. Yuan, and W. Zhao, “Bayesian Inferring Nucleon’s Gravitation Form Factors via Near-threshold J/ψPhotoproduction,”arXiv:2501.10532 [hep-ph]. [20]MAP (Multi-dimensional Analyses of Partonic distributions)Collaboration, A. Bacchetta, V. Bertone, C. Bissolotti, M. Cerut...

    2025 arXiv

  10. [21]

    Determination of unpolarized TMD distributions from the fit of Drell-Yan and SIDIS data at N 4LL,

    V. Moos, I. Scimemi, A. Vladimirov, and P. Zurita, “Determination of unpolarized TMD distributions from the fit of Drell-Yan and SIDIS data at N 4LL,” arXiv:2503.11201 [hep-ph]. [22]Jefferson Lab Angular Momentum (JAM) Collaboration, P. C. Barry, L. Gamberg, W. Melnitchouk, E....

    2023

  11. [23]

    NNPDFpol2.0: a global determination of polarised PDFs and their uncertainties at next-to-next-to-leading order,

    J. Cruz-Martinez, T. Hasenack, F. Hekhorn, G. Magni, E. R. Nocera, T. R. Rabemananjara, J. Rojo, T. Sharma, and G. van Seeventer, “NNPDFpol2.0: a global determination of polarised PDFs and their uncertainties at next-to-next-to-leading order,” arXiv:2503.11814 [hep-ph]. [24]MA...

    2022 arXiv

  12. [25]

    Colloquium: Machine learning in nuclear physics,

    A. Boehnleinet al., “Colloquium: Machine learning in nuclear physics,”Rev. Mod. Phys.94no. 3, (2022) 031003,arXiv:2112.02309 [nucl-th]

    2022 arXiv

  13. [26]

    Publishing unbinned differential cross section results,

    M. Arratiaet al., “Publishing unbinned differential cross section results,”JINST17no. 01, (2022) P01024, arXiv:2109.13243 [hep-ph]

    2022 arXiv

  14. [27]

    Unbinned inclusive cross-section measurements with machine-learned systematic uncertainties,

    L. Benato, C. Giordano, C. Krause, A. Li, R. Sch¨ ofbeck, D. Schwarz, M. Shooshtari, and D. Wang, “Unbinned inclusive cross-section measurements with machine-learned systematic uncertainties,” 5, 2025. arXiv:2505.05544 [hep-ph]

    2025

  15. [28]

    Optimal binning of X-ray spectra and response matrix design,

    J. S. Kaastra and J. A. M. Bleeker, “Optimal binning of X-ray spectra and response matrix design,”Astron. Astrophys.587(2016) A151,arXiv:1601.05309 [astro-ph.IM]

    2016 arXiv

  16. [29]

    INFERNO: Inference-Aware Neural Optimisation,

    P. De Castro and T. Dorigo, “INFERNO: Inference-Aware Neural Optimisation,”Comput. Phys. Commun.244(2019) 170–179,arXiv:1806.04743 [stat.ML]

    2019 arXiv

  17. [30]

    PDE-Foam: A probability density estimation method using self-adapting phase-space binning,

    D. Dannheim, A. Voigt, K.-J. Grahn, P. Speckmayer, and T. Carli, “PDE-Foam: A probability density estimation method using self-adapting phase-space binning,”Nucl. Instrum. Meth. A606(2009) 717–727, arXiv:0812.0922 [physics.data-an]. [31]TMV ACollaboration, A. Hockeret al., “TM...

    2009 arXiv

  18. [32]

    Constraining Effective Field Theories with Machine Learning,

    J. Brehmer, K. Cranmer, G. Louppe, and J. Pavez, “Constraining Effective Field Theories with Machine Learning,”Phys. Rev. Lett.121no. 11, (2018) 111801, arXiv:1805.00013 [hep-ph]

    2018 arXiv

  19. [33]

    The frontier of simulation-based inference,

    K. Cranmer, J. Brehmer, and G. Louppe, “The frontier of simulation-based inference,”Proc. Nat. Acad. Sci. 117no. 48, (2020) 30055–30062,arXiv:1911.01429 [stat.ML]. [34]JETSCAPECollaboration, D. Everettet al., “Multisystem Bayesian constraints on the transport coefficients of Q...

    2020 arXiv

  20. [35]

    Constraining the Higgs potential with neural simulation-based inference for di-Higgs production,

    R. Mastandrea, B. Nachman, and T. Plehn, “Constraining the Higgs potential with neural simulation-based inference for di-Higgs production,” Phys. Rev. D110no. 5, (2024) 056004, arXiv:2405.15847 [hep-ph]

    2024 arXiv

  21. [36]

    QCD Theory meets Information Theory,

    B. Assi, S. H¨ oche, K. Lee, and J. Thaler, “QCD Theory meets Information Theory,”arXiv:2501.17219 [hep-ph]

  22. [37]

    Generator Based Inference (GBI),

    C. L. Cheng, R. Das, R. Li, R. Mastandrea, V. Mikuni, B. Nachman, D. Shih, and G. Singh, “Generator Based Inference (GBI),”arXiv:2506.00119 [hep-ph]

    arXiv

  23. [38]

    Matching NLO QCD computations with Parton Shower simulations: the POWHEG method,

    S. Frixione, P. Nason, and C. Oleari, “Matching NLO QCD computations with Parton Shower simulations: the POWHEG method,”JHEP11(2007) 070, arXiv:0709.2092 [hep-ph]. [39]NNLOJETCollaboration, A. Husset al., “NNLOJET: a parton-level event generator for jet cross sections at NNLO ...

    2007 arXiv

  24. [40]

    NNLO jet production in neutral and charged current polarized deep inelastic scattering,

    I. Borsa, D. de Florian, and I. Pedron, “NNLO jet production in neutral and charged current polarized deep inelastic scattering,”Phys. Rev. D107no. 5, (2023) 054027,arXiv:2212.06625 [hep-ph]

    2023 arXiv

  25. [41]

    PARTONS: PARtonic Tomography Of Nucleon Software: A computing framework for the phenomenology of Generalized Parton Distributions,

    B. Berthouet al., “PARTONS: PARtonic Tomography Of Nucleon Software: A computing framework for the phenomenology of Generalized Parton Distributions,” Eur. Phys. J. C78no. 6, (2018) 478, arXiv:1512.06174 [hep-ph]. 14

    2018 arXiv

  26. [42]

    Study of deeply virtual Compton scattering at the future Electron-Ion Collider,

    E. C. Aschenaueret al., “Study of deeply virtual Compton scattering at the future Electron-Ion Collider,”arXiv:2503.05908 [hep-ph]

    arXiv

  27. [43]

    MILOU: A Monte-Carlo for deeply virtual Compton scattering,

    E. Perez, L. Schoeffel, and L. Favart, “MILOU: A Monte-Carlo for deeply virtual Compton scattering,” arXiv:hep-ph/0411389

    arXiv

  28. [44]

    The dipole model Monte Carlo generator Sartre 1,

    T. Toll and T. Ullrich, “The dipole model Monte Carlo generator Sartre 1,”Comput. Phys. Commun.185 (2014) 1835–1853,arXiv:1307.8059 [hep-ph]

    2014 arXiv

  29. [45]

    Exclusive vector meson production at an electron-ion collider,

    M. Lomnitz and S. Klein, “Exclusive vector meson production at an electron-ion collider,”Phys. Rev. C99 no. 1, (2019) 015203,arXiv:1803.06420 [nucl-ex]

    2019 arXiv

  30. [46]

    Υ photoproduction on the proton at the Electron-Ion Collider,

    O. Gryniuk, S. Joosten, Z.-E. Meziani, and M. Vanderhaeghen, “Υ photoproduction on the proton at the Electron-Ion Collider,”Phys. Rev. D102no. 1, (2020) 014016,arXiv:2005.09293 [hep-ph]

    2020 arXiv

  31. [47]

    Invertible Networks or Partons to Detector and Back Again,

    M. Bellagente, A. Butter, G. Kasieczka, T. Plehn, A. Rousselot, R. Winterhalder, L. Ardizzone, and U. K¨ othe, “Invertible Networks or Partons to Detector and Back Again,”SciPost Phys.9(2020) 074, arXiv:2006.06685 [hep-ph]

    2020 arXiv

  32. [48]

    Morphing parton showers with event derivatives,

    B. Nachman and S. Prestel, “Morphing parton showers with event derivatives,”arXiv:2208.02274 [hep-ph]

    arXiv

  33. [49]

    MadNIS - Neural multi-channel importance sampling,

    T. Heimel, R. Winterhalder, A. Butter, J. Isaacson, C. Krause, F. Maltoni, O. Mattelaer, and T. Plehn, “MadNIS - Neural multi-channel importance sampling,” SciPost Phys.15no. 4, (2023) 141,arXiv:2212.06172 [hep-ph]

    2023 arXiv

  34. [50]

    Interpretable deep learning models for the inference and classification of LHC data,

    V. S. Ngairangbam and M. Spannowsky, “Interpretable deep learning models for the inference and classification of LHC data,”JHEP05(2024) 004,arXiv:2312.12330 [hep-ph]

    2024 arXiv

  35. [51]

    Fitting a deep generative hadronization model,

    J. Chan, X. Ju, A. Kania, B. Nachman, V. Sangli, and A. Siodmok, “Fitting a deep generative hadronization model,”JHEP09(2023) 084,arXiv:2305.17169 [hep-ph]

    2023 arXiv

  36. [52]

    Accurate Surrogate Amplitudes with Calibrated Uncertainties,

    H. Bahl, N. Elmer, L. Favaro, M. Haußmann, T. Plehn, and R. Winterhalder, “Accurate Surrogate Amplitudes with Calibrated Uncertainties,”arXiv:2412.12069 [hep-ph]

  37. [53]

    Differentiable MadNIS-Lite,

    T. Heimel, O. Mattelaer, T. Plehn, and R. Winterhalder, “Differentiable MadNIS-Lite,”SciPost Phys.18no. 1, (2025) 017,arXiv:2408.01486 [hep-ph]

    2025 arXiv

  38. [54]

    Communicating Likelihoods with Normalising Flows,

    J. Y. Araz, A. Beck, M. Reboud, M. Spannowsky, and D. van Dyk, “Communicating Likelihoods with Normalising Flows,”arXiv:2502.09494 [hep-ph]. [55]GEANT4Collaboration, S. Agostinelliet al., “GEANT4 - A Simulation Toolkit,”Nucl. Instrum. Meth. A506(2003) 250–303

    2003 arXiv

  39. [56]

    Unfolding with Generative Adversarial Networks,

    K. Datta, D. Kar, and D. Roy, “Unfolding with Generative Adversarial Networks,”arXiv:1806.00433 [physics.data-an]

    arXiv

  40. [57]

    How to GAN away Detector Effects,

    M. Bellagente, A. Butter, G. Kasieczka, T. Plehn, and R. Winterhalder, “How to GAN away Detector Effects,” SciPost Phys.8no. 4, (2020) 070,arXiv:1912.00477 [hep-ph]

    2020 arXiv

  41. [58]

    Learning to simulate high energy particle collisions from unlabeled data,

    J. N. Howard, S. Mandt, D. Whiteson, and Y. Yang, “Learning to simulate high energy particle collisions from unlabeled data,”Sci. Rep.12(2022) 7567, arXiv:2101.08944 [hep-ph]

    2022 arXiv

  42. [59]

    Neural Empirical Bayes: Source Distribution Estimation and its Applications to Simulation-Based Inference,

    M. Vandegar, M. Kagan, A. Wehenkel, and G. Louppe, “Neural Empirical Bayes: Source Distribution Estimation and its Applications to Simulation-Based Inference,”arXiv:2011.05836 [stat.ML]

    2011 arXiv

  43. [60]

    An unfolding method based on conditional invertible neural networks (cINN) using iterative training,

    M. Backes, A. Butter, M. Dunford, and B. Malaescu, “An unfolding method based on conditional invertible neural networks (cINN) using iterative training,” SciPost Phys. Core7no. 1, (2024) 007, arXiv:2212.08674 [hep-ph]

    2024 arXiv

  44. [61]

    Unifying simulation and inference with normalizing flows,

    H. Du, C. Krause, V. Mikuni, B. Nachman, I. Pang, and D. Shih, “Unifying simulation and inference with normalizing flows,”Phys. Rev. D111no. 7, (2025) 076004,arXiv:2404.18992 [hep-ph]

    2025 arXiv

  45. [62]

    Improving generative model-based unfolding with Schr¨ odinger bridges,

    S. Diefenbacher, G.-H. Liu, V. Mikuni, B. Nachman, and W. Nie, “Improving generative model-based unfolding with Schr¨ odinger bridges,”Phys. Rev. D109 no. 7, (2024) 076011,arXiv:2308.12351 [hep-ph]

    2024

  46. [63]

    Moment extraction using an unfolding protocol without binning,

    K. Desai, B. Nachman, and J. Thaler, “Moment extraction using an unfolding protocol without binning,”Phys. Rev. D110no. 11, (2024) 116013, arXiv:2407.11284 [hep-ph]

    2024 arXiv

  47. [64]

    Generative unfolding with distribution mapping,

    A. Butter, S. Diefenbacher, N. Huetsch, V. Mikuni, B. Nachman, S. Palacios Schweitzer, and T. Plehn, “Generative unfolding with distribution mapping,” SciPost Phys.18no. 6, (2025) 200,arXiv:2411.02495 [hep-ph]

    2025 arXiv

  48. [65]

    How to Unfold Top Decays,

    L. Favaro, R. Kogler, A. Paasch, S. Palacios Schweitzer, T. Plehn, and D. Schwarz, “How to Unfold Top Decays,”arXiv:2501.12363 [hep-ph]

    arXiv

  49. [66]

    OmniFold: A Method to Simultaneously Unfold All Observables,

    A. Andreassen, P. T. Komiske, E. M. Metodiev, B. Nachman, and J. Thaler, “OmniFold: A Method to Simultaneously Unfold All Observables,”Phys. Rev. Lett.124no. 18, (2020) 182001,arXiv:1911.09107 [hep-ph]

    2020 arXiv

  50. [67]

    Scaffolding Simulations with Deep Learning for High-dimensional Deconvolution,

    A. Andreassen, P. T. Komiske, E. M. Metodiev, B. Nachman, A. Suresh, and J. Thaler, “Scaffolding Simulations with Deep Learning for High-dimensional Deconvolution,” in9th International Conference on Learning Representations. 5, 2021.arXiv:2105.04448 [stat.ML]

    2021 arXiv

  51. [68]

    SwdFold:A Reweighting and Unfolding method based on Optimal Transport Theory,

    C.-C. Pan, X. Dong, Y.-C. Sun, A.-Y. Cheng, A.-B. Wang, Y.-X. Hu, and H. Cai, “SwdFold:A Reweighting and Unfolding method based on Optimal Transport Theory,”arXiv:2406.01635 [physics.data-an]

    arXiv

  52. [69]

    Tools for unbinned unfolding,

    R. Milton, V. Mikuni, T. Lee, M. Arratia, T. Wamorkar, and B. Nachman, “Tools for unbinned unfolding,”JINST20no. 05, (2025) P05034, arXiv:2503.09720 [hep-ph]

    2025 arXiv

  53. [70]

    High-Dimensional Unfolding in Large Backgrounds,

    A. Falc˜ ao and A. Takacs, “High-Dimensional Unfolding in Large Backgrounds,”arXiv:2507.06291 [hep-ph]. [71]H1Collaboration, V. Andreevet al., “Measurement of Lepton-Jet Correlation in Deep-Inelastic Scattering with the H1 Detector Using Machine Learning for Unfolding,”Phys. R...

    2022 arXiv

  54. [74]

    CaloChallenge 2022: A Community Challenge for Fast Calorimeter Simulation,

    O. Amramet al., “CaloChallenge 2022: A Community Challenge for Fast Calorimeter Simulation,” arXiv:2410.21611 [physics.ins-det]

    2022

  55. [75]

    Unbinned Inference with Correlated Events,

    K. Desai, O. Long, and B. Nachman, “Unbinned Inference with Correlated Events,”arXiv:2504.14072 [physics.data-an]. [76]NNPDFCollaboration, R. D. Ballet al., “Parton distributions from high-precision collider data,”Eur. Phys. J. C77no. 10, (2017) 663,arXiv:1706.00428 [hep-ph]

    2017

  56. [77]

    Bayesian approach to inverse problems: an application to NNPDF closure testing,

    L. Del Debbio, T. Giani, and M. Wilson, “Bayesian approach to inverse problems: an application to NNPDF closure testing,”Eur. Phys. J. C82no. 4, (2022) 330,arXiv:2111.05787 [hep-ph]

    2022 arXiv

  57. [78]

    New generation of parton distributions with uncertainties from global QCD analysis,

    J. Pumplin, D. R. Stump, J. Huston, H. L. Lai, P. M. Nadolsky, and W. K. Tung, “New generation of parton distributions with uncertainties from global QCD analysis,”JHEP07(2002) 012,arXiv:hep-ph/0201195. [79]Jefferson Lab Angular MomentumCollaboration, N. Sato, W. Melnitchouk, ...

    2002 arXiv

  58. [80]

    First Monte Carlo analysis of fragmentation functions from single-inclusive e+e− annihilation,

    N. Sato, J. J. Ethier, W. Melnitchouk, M. Hirai, S. Kumano, and A. Accardi, “First Monte Carlo analysis of fragmentation functions from single-inclusive e+e− annihilation,”Phys. Rev. D94no. 11, (2016) 114004,arXiv:1609.00899 [hep-ph]. [81]MAP (Multi-dimensional Analyses of Par...

    2016 arXiv

  59. [82]

    Deeply virtual Compton scattering,

    X.-D. Ji, “Deeply virtual Compton scattering,”Phys. Rev. D55(1997) 7114–7125,arXiv:hep-ph/9609381

    1997 arXiv

  60. [83]

    Unraveling hadron structure with generalized parton distributions,

    A. V. Belitsky and A. V. Radyushkin, “Unraveling hadron structure with generalized parton distributions,” Phys. Rept.418(2005) 1–387,arXiv:hep-ph/0504030

    2005 arXiv

  61. [84]

    Kernel methods for evolution of generalized parton distributions,

    A. Freese, D. Adamiak, I. Clo¨ et, W. Melnitchouk, J. W. Qiu, N. Sato, and M. Zaccheddu, “Kernel methods for evolution of generalized parton distributions,”Comput. Phys. Commun.311(2025) 109552,arXiv:2412.13450 [hep-ph]

    2025 arXiv

  62. [85]

    Code for prompt numerical computation of the leading order GPD evolution,

    A. V. Vinnikov, “Code for prompt numerical computation of the leading order GPD evolution,” arXiv:hep-ph/0604248

    arXiv

  63. [86]

    Revisiting evolution equations for generalised parton distributions,

    V. Bertone, H. Dutrieux, C. Mezrag, J. M. Morgado, and H. Moutarde, “Revisiting evolution equations for generalised parton distributions,”Eur. Phys. J. C82 no. 10, (2022) 888,arXiv:2206.01412 [hep-ph]

    2022 arXiv

  64. [87]

    Deconvolution problem of deeply virtual Compton scattering,

    V. Bertone, H. Dutrieux, C. Mezrag, H. Moutarde, and P. Sznajder, “Deconvolution problem of deeply virtual Compton scattering,”Phys. Rev. D103no. 11, (2021) 114019,arXiv:2104.03836 [hep-ph]

    2021 arXiv

  65. [88]

    Shedding light on shadow generalized parton distributions,

    E. Moffat, A. Freese, I. Clo¨ et, T. Donohoe, L. Gamberg, W. Melnitchouk, A. Metz, A. Prokudin, and N. Sato, “Shedding light on shadow generalized parton distributions,”Phys. Rev. D108no. 3, (2023) 036027, arXiv:2303.12006 [hep-ph]

    2023 arXiv

  66. [89]

    Extraction of the Parton Momentum-Fraction Dependence of Generalized Parton Distributions from Exclusive Photoproduction,

    J.-W. Qiu and Z. Yu, “Extraction of the Parton Momentum-Fraction Dependence of Generalized Parton Distributions from Exclusive Photoproduction,”Phys. Rev. Lett.131no. 16, (2023) 161902, arXiv:2305.15397 [hep-ph]. [90]GEANT4Collaboration, S. Agostinelliet al., “Geant4 - A Simul...

    2023 arXiv

  67. [91]

    Flow-based generative models for Markov chain Monte Carlo in lattice field theory,

    M. S. Albergo, G. Kanwar, and P. E. Shanahan, “Flow-based generative models for Markov chain Monte Carlo in lattice field theory,”Phys. Rev. D100no. 3, (2019) 034515,arXiv:1904.12072 [hep-lat]

    2019 arXiv

  68. [92]

    Resampling base distributions of normalizing flows,

    V. Stimper, B. Sch¨ olkopf, and J. Miguel Hernandez-Lobato, “Resampling base distributions of normalizing flows,” inProceedings of The 25th International Conference on Artificial Intelligence and Statistics, G. Camps-Valls, F. J. R. Ruiz, and I. Valera, eds., vol. 151 ofProcee...

    2022

  69. [93]

    Accelerating Markov Chain Monte Carlo sampling with diffusion models,

    N. T. Hunt-Smith, W. Melnitchouk, F. Ringer, N. Sato, A. W. Thomas, and M. J. White, “Accelerating Markov Chain Monte Carlo sampling with diffusion models,” Comput. Phys. Commun.296(2024) 109059, arXiv:2309.01454 [hep-ph]

    2024 arXiv

  70. [94]

    Generative adversarial nets,

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” inAdvances in Neural Information Processing Systems, Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, and K. Weinberger, eds., vol. 27....

    2014

  71. [95]

    Sagips: A scalable asynchronous generative inverse problem solver,

    D. Lersch, M. Schram, Z. Dai, K. Rajput, X. Wu, N. Sato, and J. T. Childers, “Sagips: A scalable asynchronous generative inverse problem solver,”arXiv preprint arXiv:2407.00051(2024) . https://arxiv.org/abs/2407.00051

    2024 arXiv

  72. [96]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” inInternational Conference on Learning Representations. 2015. https://arxiv.org/abs/1412.6980

    2015 arXiv

Pith tools

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