Pith. sign in

REVIEW 4 major objections 4 minor 106 references

This paper claims that the 208Pb gluon dipole amplitude can be determined from LHC data without any assumed initial shape, yielding a saturation-scale ratio Q²_s0,Pb/Q²_s0,p = 3.17 and a lead initial condition that is well described a poste

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 04:52 UTC pith:WTCTBC2F

load-bearing objection First genuinely free-form Pb dipole extraction, but the poor coherent J/psi fit means the headline saturation-scale ratio is conditional until the flux/area systematic is quantified. the 4 major comments →

arxiv 2607.27603 v1 pith:WTCTBC2F submitted 2026-07-30 hep-ph nucl-exnucl-th

Unbiased Data-Driven Determination of the Nuclear Dipole Amplitude in the Color Glass Condensate

Si-Wei Dai , Haowu Duan , Long-Gang Pang , Guang-You Qin , Shu-Yi Wei , Han-Zhong Zhang , Wenbin Zhao This is my paper
classification hep-ph nucl-exnucl-th PACS 12.38.-t24.85.+p
keywords color glass condensategluon saturationdipole amplitudeBalitsky-Kovchegov equationphysics-informed neural networknuclear modification factorcoherent J/psi photoproductionsaturation scale
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.

Gluon saturation is expected to be strongest in heavy nuclei, but until now every extraction of the nuclear gluon dipole amplitude started from an assumed functional shape for its initial condition. This paper claims that this assumption can be dropped entirely: a physics-informed neural network is trained so the amplitude must satisfy the collinearly improved Balitsky–Kovchegov evolution equation while simultaneously describing forward-hadron nuclear modification factors and coherent J/ψ photoproduction data for lead-208. The result is an amplitude whose shape at the starting scale is learned, not imposed, together with a saturation-scale ratio Q²_s0,Pb/Q²_s0,p = 3.17 and an effective radius R_Pb ≈ 7.83 fm. The extracted lead initial condition turns out to be well described by a McLerran–Venugopalan-type form, a property that emerged from the fit rather than being assumed. If correct, this gives the saturation program a nuclear dipole amplitude derived from data rather than from model templates, ready to be evolved and used in predictions for other nuclei and collision systems.

Core claim

The central claim is that the impact-parameter-averaged dipole amplitude of 208Pb can be determined directly from experimental data within the Color Glass Condensate framework, with no parametric form imposed on its initial condition. The amplitude is represented by a neural network whose training objective includes the residual of the collinearly improved Balitsky–Kovchegov equation, so QCD evolution is enforced during training rather than checked afterward; momentum-space positivity and the black-disk limit are also imposed. Fitting the LHCb forward-hadron R_pPb data and ALICE/CMS/ATLAS/LHCb coherent J/ψ data yields Q²_s0,Pb/Q²_s0,p = 3.17^{+0.17}_{-0.10} at x_B=0.01, an effective Pb radiu

What carries the argument

The load-bearing object is the impact-parameter-averaged color dipole amplitude N(r, x_B) — the probability that a quark–antiquark dipole of size r scatters off the nucleus — whose energy evolution is governed by the collinearly improved Balitsky–Kovchegov (ciBK) equation. The paper's key move is to make this equation part of the neural-network training loss, so the amplitude is forced to satisfy QCD evolution while fitting the data, with no functional ansatz for its value at the starting scale x_0 = 0.01. The saturation scale is then defined through N(√2/Q_s0, x_0) = 1 − e^{−1/2}, and the effective transverse area S_Pb^⊥ = πR_Pb² is the only adjustable nuclear-geometry parameter entering bo

Load-bearing premise

The whole extraction rests on the assumption that one impact-parameter-averaged amplitude and a single effective transverse area S_Pb^⊥ = πR_Pb² can simultaneously describe both forward R_pPb hadron production and coherent J/ψ photoproduction, and that these two data sets are mutually consistent — an assumption the paper itself notes is strained, reporting that the J/ψ data are fit far more poorly than the R_pPb data and attributing part of the mismatch to impact-parameter-de

What would settle it

Extract the Pb dipole amplitude from R_pPb alone and from the coherent J/ψ data alone; if the two resulting Q²_s0,Pb values differ by more than the quoted uncertainty, the joint-fit premise fails. Alternatively, if a photon-flux treatment that correctly handles EMD-tagged UPC events shifts the J/ψ cross section by the approximately 20% needed to bring the saturation-scale ratio into agreement with the geometric estimate, the extracted amplitude changes accordingly.

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

If this is right

  • The same extracted Pb amplitude, evolved with ciBK, reproduces the measured R_pPb and coherent J/ψ cross sections across the available kinematic range.
  • The ratio Q²_s0,Pb/Q²_s0,p = 3.17^{+0.17}_{-0.10} provides a first model-independent quantitative test of the geometric scaling expectation A S_p^⊥/S_Pb^⊥ ≈ 2.57, exceeding it by about 20%.
  • The Pb initial condition is well described a posteriori by an MV-type form, unlike the proton, reflecting the higher color-charge density of a large nucleus.
  • Predictions of the rapidity dependence of the mean-transverse-momentum ratio in pp, pPb, and Pbp collisions agree with LHCb low-multiplicity data without system-dependent parameters.
  • Opacity-interpolated initial amplitudes for intermediate nuclei (C, O, Al, Ar, Ag, Au) provide benchmark inputs for small-x phenomenology at the LHC, RHIC, and a future electron-ion collider.

Where Pith is reading between the lines

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

  • Editorial extension: the same data-driven procedure, applied to the proton extraction with higher-statistics or future collider data, would make the proton–nucleus comparison genuinely apples-to-apples; the ratio quoted here inherits any bias in the proton amplitude used as baseline.
  • Editorial extension: since the paper's own discussion implies the J/ψ channel is subject to impact-parameter-dependent systematic shifts, the extracted amplitude's saturation scale should shift if a photon-flux treatment consistent across electromagnetic-dissociation-tagged measurements is adopted — a testable consequence the authors themselves flag.
  • Editorial extension: the opacity-scaling interpolation to intermediate nuclei provides ready-made initial conditions; a natural next check is exclusive J/ψ or ρ photoproduction off those nuclei, where the central prediction is a smooth A-ordering of saturation scales.

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 / 4 minor

Summary. The paper presents a physics-informed neural network (PINN) determination of the impact-parameter-averaged dipole amplitude N(r,x) for 208Pb. The network is trained to satisfy the collinearly improved BK equation while simultaneously constrained by LHCb forward R_pPb data and coherent J/ψ photoproduction in Pb–Pb ultraperipheral collisions; no functional form for the nuclear initial condition at the starting scale is imposed. The authors report a good R_pPb fit (χ²/N = 0.446), a poor coherent J/ψ fit (χ²/N = 6.228), an effective Pb radius R_Pb = 7.83 fm, a saturation-scale ratio Q_{s0,Pb}^2/Q_{s0,p}^2 = 3.17, an a posteriori MV-type description of the Pb initial condition, and out-of-sample predictions for the mean-transverse-momentum ratio in pp, pPb, and Pbp collisions compared with preliminary LHCb data. They also interpolate to other nuclei via opacity scaling.

Significance. If the extraction is reliable, this is a significant methodological advance: it replaces parametric nuclear initial conditions with a flexible, evolution-constrained representation and provides a data-driven nuclear dipole amplitude. The paper contains several valuable cross-checks: an independent ciBK solver reproduces the PINN solution; momentum-space Fourier positivity is verified; R_pPb is shown to be stable against fragmentation-function and factorization-scale choices; and the LHCb ⟨pT⟩ comparison is genuinely out-of-sample. However, the coherent J/ψ sector is not described at the quoted precision, and because the same N_Pb and S_Pb enter both observables, the headline Q_s0 ratio absorbs the discrepancy without a propagated systematic. The central quantitative claims therefore need revision before the paper can be accepted; the methodology is promising and the issues are localizable and fixable.

major comments (4)
  1. [§III.A, Fig. 3(a); Eqs. (4), (8), (31)] The joint fit gives χ²/N = 6.228 for coherent J/ψ while χ²/N = 0.446 for R_pPb, yet the Conclusion states that the amplitude 'reproduces the measured R_pPb and coherent J/ψ cross sections.' Because the same N_Pb and S_Pb^⊥ enter both observables (Eqs. 4 and 8), the extracted amplitude is a compromise between data sets that are inconsistent under the model. The EMD-tagged flux caveat identifies an impact-parameter-dependent systematic, but it is not propagated into R_Pb or into the uncertainty of Eq. (31). Please either (i) add an EMD-flux/area uncertainty parameter, (ii) refit without the J/ψ data and show that Eq. (31) is stable, or (iii) demote the J/ψ data to a consistency check. Without this, the 'unbiased extraction' claim is overstated.
  2. [Abstract; §III.B, Eqs. (31)–(32)] The paper calls Q_{s0,Pb}^2/Q_{s0,p}^2 = 3.17^{+0.17}_{-0.10} 'consistent with simple geometric scaling', but the geometric estimate is 2.57^{+0.08}_{-0.10}. This is a ~20% difference, roughly 3–4σ on the quoted errors. The sentence 'well within the accuracy expected of Eq. (32)' is an assertion because the theoretical uncertainty of Eq. (32) is not quantified. Either estimate that uncertainty (e.g., from impact-parameter fluctuations and color-charge fluctuations) or describe the comparison as qualitative agreement. This is a headline claim and cannot rest on an unquantified 'expected accuracy'.
  3. [§III.B and §IV (geometric-scaling 'test')] The geometric-scaling comparison is not an independent test: both the left side of Eq. (31) and the radius ratio in Eq. (32) are outputs of the same joint fit. A 20% offset could be absorbed by the fitted S_Pb^⊥. The Conclusion's phrase 'a quantitative test of the accuracy of this widely used geometric scaling argument in a model-independent setting' should be softened to 'a self-consistency check under the model'. A radius fixed by nuclear density or by t-dependent coherent diffraction would make the test genuinely independent.
  4. [§II.A and §III.B (initial rapidity)] Eq. (2) defines Y = ln(x_c/x) with x_c = 0.03, while the text repeatedly states that the initial condition is imposed at x_0 = 0.01. If x_c is the starting point, then x_0 = 0.01 is about one unit of evolution later and the 'initial condition' at x_0 is not the unparametrized network output. Please clarify whether the initial condition is at Y = 0 (x = 0.03) or at Y = ln 3 ≈ 1.1 (x = 0.01). This also affects the definition of Q_{s0,Pb} in Eq. (30) and the MV fit in Eq. (29).
minor comments (4)
  1. [§II.B and figure captions] The captions of Figs. 2 and 3 use 'χ²/d.o.f.' while the text uses χ²/N with N the number of data points. Since the network has many effective parameters, please use χ²/N_{data} consistently.
  2. [§II.B, Eq. (8)] The phrase 'kept outside in the Eq. (8)' is grammatically unclear. Rephrase to state that S_T^⊥ is pulled out of the amplitude and multiplies the sum in Eq. (8).
  3. [§III.A and §III.C] Minor typos and notation: 'R_pP b' should be 'R_pPb'; Fig. 4's lower panel uses 'xB' instead of 'x_B'; Fig. 6 compares against 'LHCb preliminary' data from Ref. [56], which should be explicitly flagged as preliminary and not final in the main text.
  4. [Ref. [92] and data availability] Ref. [92] is listed as 'To appear in zenodo soon'. The tabulated dipole amplitudes and their Fourier transforms are central outputs of the paper; the repository should be finalized and referenced before publication.

Circularity Check

0 steps flagged

No significant circularity: the extraction is data-driven, with out-of-sample LHCb predictions and externally validated proton inputs.

full rationale

The paper's derivation chain is not circular. The Pb dipole amplitude N_Pb is obtained by minimizing a physics-informed loss that includes the ciBK residual and two data classes that are not derived from each other: forward R_pPb and coherent J/psi photoproduction. No fitted parameter is renamed as a prediction: K_h cancels in R_pPb, K_V = 1.079 is fixed from a previous ep analysis, and the only adjustable nuclear-geometry parameter is the transverse area S_Pb^perp = pi R_Pb^2, which is fitted to the same observables rather than used as a prediction. The geometric-scaling 'test' comparing Eq. (31) with Eq. (32) does compare two outputs of the extraction, so it is more an internal consistency check than an independent test; however, no equation forces Eq. (31) to equal Eq. (32), and the comparison is not an identity or a renamed fit parameter, so it is not circular by construction. The LHCb mean-transverse-momentum ratio was not used in training and constitutes a genuine out-of-sample prediction; the pp/RHIC comparisons in the appendices provide external validation of the proton inputs taken from the authors' previous work. The large chi^2/N = 6.228 for the coherent J/psi cross section (Fig. 3a) and the authors' discussion of EMD-tagged photon-flux ambiguities indicate a data-model tension and an impact-parameter modeling limitation, but this is an accuracy/systematic issue, not a circularity of the derivation chain.

Axiom & Free-Parameter Ledger

6 free parameters · 6 axioms · 0 invented entities

The ledger contains (a) fitted or inherited numerical inputs the extraction depends on — R_Pb is the only parameter fitted anew here, while K_V, C, α_fr, α, and Q0 are carried from prior work — and (b) modeling premises of the dipole-CGC framework that define what 'unbiased' can mean. No new physical particles, forces, or dynamical entities are introduced; the neural network and learned variance field are computational constructs.

free parameters (6)
  • R_Pb (effective transverse radius of 208Pb) = 7.83^{+0.15}_{-0.12} fm
    The only nuclear-geometry parameter adjusted in this work; sets S_Pb⊥ = πR_Pb² entering both the forward-hadron cross section (Eq. 4) and J/ψ photoproduction (Eq. 8). Fitted to the combined data; larger than the ~5.5 fm charge radius, absorbing normalization uncertainties.
  • K_V (J/ψ wave-function normalization) = 1.079
    Fixed from the authors' previous proton extraction [35], not fitted here; scales the entire coherent J/ψ cross section (Eq. 8). A wrong K_V shifts the extracted amplitude's normalization.
  • C (scale constant in BLM running coupling) = not quoted; fixed from prior fit
    Appendix A: 'the dimensionless constant C ... is treated as a free fit parameter' in the ciBK kernel (Eqs. A5–A6); carried over from the proton fit rather than re-fitted here.
  • α_fr (frozen infrared coupling) = 0.7
    Chosen by hand following Ref. [29]; regulates the Landau pole in the running coupling (Eq. A6) and affects the evolution speed.
  • α (factorization/fragmentation scale multiplier) = 3
    μ² = α²(μ²_min + pT²), following Ref. [55]; Appendix C shows R_pA is weakly sensitive to α ∈ {2,3,4}, so its impact is mild.
  • Q0 (dimensionful scale in the network output) = 1 GeV²
    Introduced in Eq. (13) to make the exponential dimensionless; fixed convention, absorbable into network weights.
axioms (6)
  • domain assumption The collinearly-improved BK equation (Eq. 2) with rapidity shifts δ_i, STL coefficient A_1 = 11/12, and the DLA factor truncated to K_DLA = 1 is the correct evolution law for the b-averaged dipole amplitude.
    Central dynamical input; if the kernel is wrong, the network's evolved amplitude is biased. The K_DLA = 1 truncation is justified in Appendix A by the claim that rapidity shifts generate 'exactly the same double-logarithmic series', and partly by stability ('keeps the evolution speed stable').
  • domain assumption Leading-order hybrid factorization (Eq. 4) with a single K_h that cancels exactly in R_pA describes forward hadron production.
    All R_pPb constraints flow through this LO formula; the quantitative reach (pT up to ~8 GeV) relies on the K-factor cancellation being exact.
  • domain assumption A single impact-parameter-averaged amplitude and one effective transverse area S_Pb⊥ describe both forward hadron production and coherent J/ψ photoproduction.
    Stated in §II.B; if UPC photon-flux systematics (EMD tagging) shift the effective b-range for J/ψ relative to pPb hadron production, the joint fit is biased. The authors flag this as a possible source of the χ²/N = 6.2 tension (Fig. 3a).
  • domain assumption The boosted-Gaussian J/ψ wave function, skewness R_g, and real-part β corrections (Appendix D) are adequate.
    Determines the mapping from N(r) to σ(γA→J/ψ A); wave-function uncertainties are folded into K_V, fixed from Ref. [35].
  • domain assumption STARlight photon flux n_A(ω) with the symmetric-UPC separation (Eq. 9) describes the rapidity distribution.
    Converts σ(W) into dσ/dy; the EMD-tagging ambiguity raised by Refs. [81,82] is acknowledged by the authors.
  • ad hoc to paper The network parameterization (Eq. 13) is flexible enough to represent the true amplitude, and the gradient-calibrated, manually set loss weights (w_ciBK = 5×10^5 dominating) yield the correct PDE-data balance.
    The extracted initial condition is only 'unbiased' if the architecture imposes no significant shape bias and the manually set weights do not over- or under-enforce the evolution residual.

pith-pipeline@v1.3.0-daily-deepseek · 23722 in / 28480 out tokens · 309284 ms · 2026-08-01T04:52:41.401343+00:00 · methodology

0 comments
read the original abstract

Gluon saturation limits the growth of parton densities at small Bjorken-$x$ and is expected to be most pronounced in heavy nuclei. Yet quantitative extractions of the nuclear gluon dipole amplitude have long relied on parametrized initial conditions, introducing uncontrolled model dependence that obscures genuine nuclear effects. We introduce a physics-informed neural-network framework that embeds the collinearly improved Balitsky-Kovchegov evolution equation directly into the training objective, allowing the impact-parameter-averaged dipole amplitude to be determined from data without assuming a functional form for its initial condition. Applying this framework to forward-hadron nuclear-modification-factor and coherent $J/\psi$ photoproduction data, we extract the $^{208}$Pb dipole amplitude at $x_0=0.01$ with QCD evolution and momentum-space positivity enforced throughout training. The evolved amplitude reproduces the measured cross sections across the available kinematic range and yields a saturation-scale ratio $Q_{s0,\mathrm{Pb}}^2/Q_{s0,p}^2 = 3.17^{+0.17}_{-0.10}$, consistent with simple geometric scaling. The extracted Pb initial condition is well described by a McLerran-Venugopalan-type form, in contrast to the proton, reflecting the higher color-charge density of a large nucleus. Using the same amplitude, we predict the rapidity dependence of the transverse-momentum ratio in $pp$, $p$Pb, and Pb$p$ collisions, finding agreement with recent LHCb measurements at low multiplicity without any system-dependent parameters. This work provides the first unbiased, data-driven determination of nuclear structure in the saturation regime and establishes a general strategy for embedding nonlinear evolution equations into machine-learning extractions of dynamically constrained observables.

Figures

Figures reproduced from arXiv: 2607.27603 by Guang-You Qin, Han-Zhong Zhang, Haowu Duan, Long-Gang Pang, Shu-yi Wei, Si-Wei Dai, Wenbin Zhao.

Figure 1
Figure 1. Figure 1: FIG. 1. Architecture of the physics-informed neural network used to extract the nuclear dipole amplitude. Given ( [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: shows RpPb at √ sNN = 5.02 TeV evaluated with three different fragmentation-function sets, com￾pared to the LHCb measurements across several rapidity bins. Our PINN extraction reproduces the data well over￾all, and the agreement is especially good at the relatively low rapidities 2 < y < 4. This is consistent with Ref. [55] that found that RpPb in this range is largely insensitive to higher-order contribut… view at source ↗
Figure 3
Figure 3. Figure 3: shows the coherent J/ψ photoproduction cross section in ultraperipheral Pb–Pb collisions at √ sNN = 5.02 and 5.36 TeV, as a function of W and of rapidity, respectively. Compared to the ALICE and CMS data, our global PINN fit slightly underestimates the cross section at low W and overestimates it at high W. 10 2 10 3 W [GeV] 10 −2 10 −1 σ(γ+ Pb→ J/ψ) [mb] Pb+Pb (UPC) χ 2 /N=6.228 PINN 1σ uncertainty ALICE 5… view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Momentum-space dipole distribution [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Extracted nuclear dipole amplitude [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Ratios of the mean transverse momentum [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: shows the resulting initial dipole amplitudes N(r, xB = 0.01) for the proton and for 12C, 16O, 27Al, 40Ar, 108Ag, 197Au, and 208Pb. The profiles are smoothly ordered in A (increasing from right to left). Starting from the unbiased, data-driven initial amplitudes for the proton and 208Pb, we construct intermediate nu￾clei through the opacity interpolation described above, and then evolve all systems in xB u… view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Differential cross sections for forward single-inclusive hadron production in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Differential cross sections for forward single-inclusive hadron production in [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Nuclear modification factor [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] 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

106 extracted references · 75 linked inside Pith

  1. [1]

    Kinematic constraints (rapidity shifts). Consistency of the small-xevolution requires that successive soft-gluon emissions be ordered in lifetime; this is implemented by shifting the rapidity arguments of the amplitudes appearing in the nonlinear term, δi = max{0,ln r2 r2 i }, i= 1,2,(A2) withrthe parent-dipole size andr 1,r 2 the sizes of the two daughte...

  2. [2]

    Double logarithmic approximation factor An alternative route to resumming the same double collinear logarithms, used in formulations that dispense with explicit rapidity shifts, is to insert the closed-form Bessel factor KDLA(ρ) = J1(2 p ¯αsρ2)p ¯αsρ2 = 1− ¯αsρ2 2 + ( ¯αsρ2)2 12 +· · ·,(A3) withρ= p Lr1r Lr2r andL rir = ln(r 2 i /r2). Expanding the rapidi...

  3. [3]

    Single transverse logarithmic correction Besides the double logarithms discussed above, the NLO corrections to the BK kernel contain single transverse logarithms, ln(r 2/r2 i ), that lie outside the reach of the time-ordering constraints. All orders of this logarithm are resummed through the STL factor KSTL = r2 min(r2 1, r2 2) ±A1 ¯αBLM , A 1 = 11 12 ,(A...

  4. [4]

    Combined with the kinematic constraints of (i) and the truncated DLA factor of (ii), this term completes the collinearly improved kernel used throughout this work

    and negative forr 2 >min(r 2 1, r2 2), ensuring that the kernel is uniformly suppressed in configurations with strongly ordered dipole sizes. Combined with the kinematic constraints of (i) and the truncated DLA factor of (ii), this term completes the collinearly improved kernel used throughout this work

  5. [5]

    fast apparent convergence

    Running-coupling prescription Among the NLO corrections to the BK equation, those enhanced by the one-loopβ-function — the running- coupling corrections — are both sizable and theoretically well established. Several schemes for incorporating them exist in the literature; we adopt the Brodsky–Lepage–Mackenzie (BLM) prescription, also known as “fast apparen...

  6. [6]

    L. V. Gribov, E. M. Levin, and M. G. Ryskin, Phys. Rept.100, 1 (1983)

    1983

  7. [7]

    A. H. Mueller and J.-w. Qiu, Nucl. Phys. B268, 427 (1986)

    1986

  8. [8]

    Y. V. Kovchegov and E. Levin,Quantum Chromody- namics at High Energy, Vol. 33 (Oxford University Press, 2013)

    2013

  9. [9]

    McLerran, Nucl

    L. McLerran, Nucl. Phys. A702, 49 (2002)

    2002

  10. [10]

    The Color glass con- densate and high-energy scattering in QCD,

    E. Iancu and R. Venugopalan, “The Color glass con- densate and high-energy scattering in QCD,” inQuark- gluon plasma 4, edited by R. C. Hwa and X.-N. Wang (2003) pp. 249–3363, arXiv:hep-ph/0303204

    Pith/arXiv arXiv 2003

  11. [11]

    Weigert, Prog

    H. Weigert, Prog. Part. Nucl. Phys.55, 461 (2005), arXiv:hep-ph/0501087

    Pith/arXiv arXiv 2005

  12. [12]

    Gelis, E

    F. Gelis, E. Iancu, J. Jalilian-Marian, and R. Venu- gopalan, Ann. Rev. Nucl. Part. Sci.60, 463 (2010), arXiv:1002.0333 [hep-ph]

    Pith/arXiv arXiv 2010

  13. [13]

    J. L. Albacete and C. Marquet, Prog. Part. Nucl. Phys. 76, 1 (2014), arXiv:1401.4866 [hep-ph]

    Pith/arXiv arXiv 2014

  14. [14]

    Morreale and F

    A. Morreale and F. Salazar, Universe7, 312 (2021), arXiv:2108.08254 [hep-ph]

    Pith/arXiv arXiv 2021

  15. [15]

    L. D. McLerran and R. Venugopalan, Phys. Rev. D49, 2233 (1994), arXiv:hep-ph/9309289

    Pith/arXiv arXiv 1994

  16. [16]

    L. D. McLerran and R. Venugopalan, Phys. Rev. D49, 3352 (1994), arXiv:hep-ph/9311205

    Pith/arXiv arXiv 1994

  17. [17]

    L. D. McLerran and R. Venugopalan, Phys. Rev. D50, 2225 (1994), arXiv:hep-ph/9402335

    Pith/arXiv arXiv 1994

  18. [18]

    Abdul Khaleket al., Nucl

    R. Abdul Khaleket al., Nucl. Phys. A1026, 122447 (2022), arXiv:2103.05419 [physics.ins-det]

    Pith/arXiv arXiv 2022

  19. [19]

    E. C. Aschenauer, S. Fazio, J. H. Lee, H. Mantysaari, B. S. Page, B. Schenke, T. Ullrich, R. Venugopalan, and P. Zurita, Rept. Prog. Phys.82, 024301 (2019), arXiv:1708.01527 [nucl-ex]

    Pith/arXiv arXiv 2019

  20. [20]

    Accardiet al., Eur

    A. Accardiet al., Eur. Phys. J. A52, 268 (2016), arXiv:1212.1701 [nucl-ex]

    Pith/arXiv arXiv 2016

  21. [21]

    C. A. Bertulani, S. R. Klein, and J. Nystrand, Ann. Rev. Nucl. Part. Sci.55, 271 (2005), arXiv:nucl- ex/0502005

    arXiv 2005

  22. [22]

    S. R. Klein and H. M¨ antysaari, Nature Rev. Phys.1, 662 (2019), arXiv:1910.10858 [hep-ex]

    Pith/arXiv arXiv 2019

  23. [23]

    D. P. Anderleet al., Front. Phys. (Beijing)16, 64701 (2021), arXiv:2102.09222 [nucl-ex]

    Pith/arXiv arXiv 2021

  24. [24]

    Arleoet al., Phys

    F. Arleoet al., Phys. Rev. C113, 040501 (2026), arXiv:2506.17454 [hep-ph]

    Pith/arXiv arXiv 2026

  25. [25]

    Balitsky, Nucl

    I. Balitsky, Nucl. Phys. B463, 99 (1996), arXiv:hep- ph/9509348

    arXiv 1996

  26. [26]

    Y. V. Kovchegov, Phys. Rev. D60, 034008 (1999), arXiv:hep-ph/9901281

    Pith/arXiv arXiv 1999

  27. [27]

    Jalilian-Marian, A

    J. Jalilian-Marian, A. Kovner, L. D. McLerran, and H. Weigert, Phys. Rev. D55, 5414 (1997), arXiv:hep- ph/9606337

    arXiv 1997

  28. [28]

    Jalilian-Marian, A

    J. Jalilian-Marian, A. Kovner, A. Leonidov, and H. Weigert, Nucl. Phys. B504, 415 (1997), arXiv:hep- ph/9701284

    arXiv 1997

  29. [29]

    Jalilian-Marian, A

    J. Jalilian-Marian, A. Kovner, A. Leonidov, and H. Weigert, Phys. Rev. D59, 014014 (1998), arXiv:hep- ph/9706377

    arXiv 1998

  30. [30]

    Iancu and L

    E. Iancu and L. D. McLerran, Phys. Lett. B510, 145 (2001), arXiv:hep-ph/0103032

    Pith/arXiv arXiv 2001

  31. [31]

    Ferreiro, E

    E. Ferreiro, E. Iancu, A. Leonidov, and L. McLerran, Nucl. Phys. A703, 489 (2002), arXiv:hep-ph/0109115

    Pith/arXiv arXiv 2002

  32. [32]

    Iancu, A

    E. Iancu, A. Leonidov, and L. D. McLerran, Phys. Lett. B510, 133 (2001), arXiv:hep-ph/0102009

    Pith/arXiv arXiv 2001

  33. [33]

    Iancu, A

    E. Iancu, A. Leonidov, and L. D. McLerran, Nucl. Phys. A692, 583 (2001), arXiv:hep-ph/0011241

    Pith/arXiv arXiv 2001

  34. [34]

    J. L. Albacete, N. Armesto, J. G. Milhano, P. Quiroga- Arias, and C. A. Salgado, Eur. Phys. J. C71, 1705 (2011), arXiv:1012.4408 [hep-ph]

    Pith/arXiv arXiv 2011

  35. [35]

    M¨ antysaari and B

    H. M¨ antysaari and B. Schenke, Phys. Rev. D98, 034013 (2018), arXiv:1806.06783 [hep-ph]

    Pith/arXiv arXiv 2018

  36. [36]

    Duclou´ e, E

    B. Duclou´ e, E. Iancu, G. Soyez, and D. N. Tri- antafyllopoulos, Phys. Lett. B803, 135305 (2020), arXiv:1912.09196 [hep-ph]

    Pith/arXiv arXiv 2020

  37. [37]

    Casuga, H

    C. Casuga, H. H¨ anninen, and H. M¨ antysaari, Phys. Rev. D112, 034003 (2025), arXiv:2506.00487 [hep-ph]

    Pith/arXiv arXiv 2025

  38. [38]

    Casuga and H

    C. Casuga and H. M¨ antysaari, (2026), arXiv:2604.22332 [hep-ph]

    Pith/arXiv arXiv 2026

  39. [39]

    Korcyl, T

    P. Korcyl, T. M. H. Le, F. Salazar, and T. Stebel, (2026), arXiv:2607.23485 [hep-ph]

    Pith/arXiv arXiv 2026

  40. [40]

    Dai, F.-P

    S.-W. Dai, F.-P. Li, L.-G. Pang, G.-Y. Qin, S.-Y. Wei, H.-Z. Zhang, and W. Zhao, (2026), arXiv:2603.08008 [hep-ph]

    Pith/arXiv arXiv 2026

  41. [41]

    Kowalski, T

    H. Kowalski, T. Lappi, and R. Venugopalan, Phys. Rev. Lett.100, 022303 (2008), arXiv:0705.3047 [hep-ph]

    Pith/arXiv arXiv 2008

  42. [42]

    Deganutti, C

    F. Deganutti, C. Royon, and S. Schlichting, JHEP01, 159 (2024), arXiv:2311.01965 [hep-ph]

    Pith/arXiv arXiv 2024

  43. [43]

    Caucal, Z.-B

    P. Caucal, Z.-B. Kang, P. Korcyl, F. Salazar, B. Schenke, T. Stebel, R. Venugopalan, and W. Zhao, Phys. Lett. B879, 140599 (2026), arXiv:2512.21466 [hep-ph]

    arXiv 2026

  44. [44]

    Z.-B. Kang, R. Kao, M. Li, and J. Penttala, Phys. Rev. D112, 076006 (2025), arXiv:2504.00069 [hep-ph]

    Pith/arXiv arXiv 2025

  45. [45]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. Karniadakis, Journal of Computational Physics378, 686 (2019)

    2019

  46. [46]

    von Rueden, S

    L. von Rueden, S. Mayer, K. Beckh, B. Georgiev, S. Giesselbach, R. Heese, B. Kirsch, J. Pfrommer, A. Pick, R. Ramamurthy, M. Walczak, J. Garcke, C. Bauckhage, and J. Schuecker, IEEE Transactions on Knowledge and Data Engineering35, 614 (2023)

    2023

  47. [47]

    Cuomo, V

    S. Cuomo, V. S. D. Cola, F. Giampaolo, G. Rozza, M. Raissi, and F. Piccialli, CoRRabs/2201.05624 (2022), 2201.05624

    Pith/arXiv arXiv 2022

  48. [48]

    Dai, F.-P

    S.-W. Dai, F.-P. Li, L.-G. Pang, X.-N. Wang, B.-W. Zhang, and H.-Z. Zhang, (2026), arXiv:2601.20177 [hep-ph]

    arXiv 2026

  49. [49]

    Kou and X

    W. Kou and X. Chen, Phys. Lett. B877, 140507 (2026), arXiv:2601.16391 [hep-ph]

    Pith/arXiv arXiv 2026

  50. [50]

    T. I. Baihaqi, C. Setyadi, Z. Akbar, P. T. P. Hutauruk, and A. Salim Adam, (2025), 10.1142/S021773232642006X, arXiv:2512.21704 [hep- ph]

    Pith/arXiv arXiv 2025

  51. [51]

    Li, L.-G

    F.-P. Li, L.-G. Pang, and G.-Y. Qin, (2026), arXiv:2604.22352 [nucl-th]

    Pith/arXiv arXiv 2026

  52. [52]

    Matsuda, K

    H. Matsuda, K. Hattori, and K. Murase, (2025), arXiv:2512.17971 [physics.flu-dyn]

    arXiv 2025

  53. [53]

    M. G. Ryskin, Z. Phys. C57, 89 (1993)

    1993

  54. [54]

    Iancu, A

    E. Iancu, A. H. Mueller, and D. N. Triantafyllopoulos, Phys. Rev. Lett.128, 202001 (2022), arXiv:2112.06353 [hep-ph]

    Pith/arXiv arXiv 2022

  55. [55]

    M¨ antysaari, B

    H. M¨ antysaari, B. Schenke, C. Shen, and W. Zhao, Phys. Rev. Lett.131, 062301 (2023), arXiv:2303.04866 [nucl-th]. 19

    Pith/arXiv arXiv 2023

  56. [56]

    Cassar, Z

    K. Cassar, Z. Wang, X. Chu, and E.-C. Aschenauer, Phys. Rev. D112, 034034 (2025), arXiv:2503.08447 [hep-ph]

    Pith/arXiv arXiv 2025

  57. [57]

    Fujii, T

    H. Fujii, T. Hirano, K. Itakura, Y. Nara, and S. Zhao, (2026), arXiv:2605.15494 [hep-ph]

    Pith/arXiv arXiv 2026

  58. [58]

    G. A. Chirilli, B.-W. Xiao, and F. Yuan, Phys. Rev. Lett.108, 122301 (2012), arXiv:1112.1061 [hep-ph]

    Pith/arXiv arXiv 2012

  59. [59]

    G. A. Chirilli, B.-W. Xiao, and F. Yuan, Phys. Rev. D 86, 054005 (2012), arXiv:1203.6139 [hep-ph]

    Pith/arXiv arXiv 2012

  60. [60]

    Y. Shi, L. Wang, S.-Y. Wei, and B.-W. Xiao, Phys. Rev. Lett.128, 202302 (2022), arXiv:2112.06975 [hep-ph]

    Pith/arXiv arXiv 2022

  61. [61]

    LHCb Overview,

    ´O. B. Garc ´ ıa, “LHCb Overview,” Presented at the 13th International Conference on Hard and Electromagnetic Probes of High-Energy Nuclear Collisions (Hard Probes 2026)

    2026

  62. [62]

    Iancu, J

    E. Iancu, J. D. Madrigal, A. H. Mueller, G. Soyez, and D. N. Triantafyllopoulos, Phys. Lett. B744, 293 (2015), arXiv:1502.05642 [hep-ph]

    Pith/arXiv arXiv 2015

  63. [63]

    Lappi and H

    T. Lappi and H. M¨ antysaari, Phys. Rev. D93, 094004 (2016), arXiv:1601.06598 [hep-ph]

    Pith/arXiv arXiv 2016

  64. [64]

    Beuf, Phys

    G. Beuf, Phys. Rev. D89, 074039 (2014), arXiv:1401.0313 [hep-ph]

    Pith/arXiv arXiv 2014

  65. [65]

    Iancu, J

    E. Iancu, J. D. Madrigal, A. H. Mueller, G. Soyez, and D. N. Triantafyllopoulos, Phys. Lett. B750, 643 (2015), arXiv:1507.03651 [hep-ph]

    Pith/arXiv arXiv 2015

  66. [66]

    Duclou´ e, E

    B. Duclou´ e, E. Iancu, A. H. Mueller, G. Soyez, and D. N. Triantafyllopoulos, JHEP04, 081 (2019), arXiv:1902.06637 [hep-ph]

    Pith/arXiv arXiv 2019

  67. [67]

    Dumitru, A

    A. Dumitru, A. Hayashigaki, and J. Jalilian-Marian, Nucl. Phys. A765, 464 (2006), arXiv:hep-ph/0506308

    Pith/arXiv arXiv 2006

  68. [68]

    A. H. Rezaeian, Phys. Lett. B718, 1058 (2013), arXiv:1210.2385 [hep-ph]

    Pith/arXiv arXiv 2013

  69. [69]

    Kowalski, L

    H. Kowalski, L. Motyka, and G. Watt, Phys. Rev. D 74, 074016 (2006), arXiv:hep-ph/0606272

    Pith/arXiv arXiv 2006

  70. [70]

    S. R. Klein, J. Nystrand, J. Seger, Y. Gorbunov, and J. Butterworth, Comput. Phys. Commun.212, 258 (2017), arXiv:1607.03838 [hep-ph]

    Pith/arXiv arXiv 2017

  71. [71]

    Loshchilov and F

    I. Loshchilov and F. Hutter, inInternational Conference on Learning Representations(2019)

    2019

  72. [72]

    Loshchilov and F

    I. Loshchilov and F. Hutter, CoRRabs/1608.03983 (2016), 1608.03983

    Pith/arXiv arXiv 2016

  73. [73]

    S. Wang, Y. Teng, and P. Perdikaris, SIAM Journal on Scientific Computing43, A3055 (2021)

    2021

  74. [74]

    Y. Wang, Y. Yao, J. Guo, and Z. Gao, Journal of Com- putational Physics510, 113112 (2024)

    2024

  75. [75]

    Z. Chen, V. Badrinarayanan, C. Lee, and A. Rabi- novich, CoRRabs/1711.02257(2017), 1711.02257

    Pith/arXiv arXiv 2017

  76. [76]

    Lakshminarayanan, A

    B. Lakshminarayanan, A. Pritzel, and C. Blundell, (2017), arXiv:1612.01474 [stat.ML]

    Pith/arXiv arXiv 2017

  77. [77]

    de Florian, R

    D. de Florian, R. Sassot, and M. Stratmann, Phys. Rev. D75, 114010 (2007), arXiv:hep-ph/0703242

    Pith/arXiv arXiv 2007

  78. [78]

    J. Gao, C. Liu, X. Shen, H. Xing, and Y. Zhao, Phys. Rev. Lett.132, 261903 (2024), arXiv:2401.02781 [hep- ph]

    Pith/arXiv arXiv 2024

  79. [79]

    J. Gao, C. Liu, X. Shen, H. Xing, and Y. Zhao, Phys. Rev. D110, 114019 (2024), arXiv:2407.04422 [hep-ph]

    Pith/arXiv arXiv 2024

  80. [80]

    Anderson, W

    T. Anderson, W. Melnitchouk, and N. Sato (JAM Col- laboration (PDF Analysis Group)), Phys. Rev. D112, 094011 (2025), arXiv:2501.00665 [hep-ph]

    arXiv 2025

Showing first 80 references.