pith. machine review for the scientific record. sign in

arxiv: 2604.22352 · v1 · submitted 2026-04-24 · ⚛️ nucl-th · hep-ph

Recognition: unknown

Four-dimensional QCD equation of state from a quasi-parton model with physics-informed neural networks

Fu-Peng Li, Guang-You Qin, Long-Gang Pang

Pith reviewed 2026-05-08 09:22 UTC · model grok-4.3

classification ⚛️ nucl-th hep-ph
keywords QCD equation of statequasi-particle modelphysics-informed neural networkslattice QCDheavy-ion collisionschemical potentialsbaryon-strangeness correlationcumulants
0
0 comments X

The pith

A physics-informed neural network quasi-parton model constructs a four-dimensional QCD equation of state from zero-density lattice data.

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

This paper develops a deep-learning-assisted quasi-particle model to generate an equation of state for QCD matter that depends on temperature together with baryon, charge, and strangeness chemical potentials. The masses of light quarks, strange quarks, and gluons are written as functions of these four variables and optimized inside a physics-informed neural network that enforces thermodynamic consistency. Training uses only lattice QCD results at vanishing chemical potentials, after which the model extrapolates to finite densities. A reader would care because hydrodynamic simulations of heavy-ion collisions require precisely this kind of four-dimensional EoS to follow the evolution of matter created at different collision energies.

Core claim

The DLQPM accurately reproduces the lattice-calculated cumulants χ^{B,Q,S}_{i,j,k} at μ_{B,Q,S}=0, and its predicted EoS at various chemical potentials agrees well with results from the generalized T'-expansion method in lattice QCD. The calculated baryon-strangeness correlation C_{BS} is consistent, within uncertainties, with preliminary STAR data.

What carries the argument

The deep-learning-assisted quasi-particle model (DLQPM) inside a physics-informed neural network, where the masses of light quarks, strange quarks, and gluons are parameterized as functions of T, μ_B, μ_Q, and μ_S.

Load-bearing premise

Quasi-particle masses can be written as functions of temperature and the three chemical potentials and trained solely on zero-chemical-potential lattice data to give thermodynamically consistent results at finite densities.

What would settle it

Lattice QCD calculations of pressure, energy density, or cumulants at small but nonzero chemical potentials that differ markedly from the DLQPM predictions would show the extrapolation has failed.

Figures

Figures reproduced from arXiv: 2604.22352 by Fu-Peng Li, Guang-You Qin, Long-Gang Pang.

Figure 1
Figure 1. Figure 1: FIG. 1. (Color online) The input layer comprises four dimensions, specifically view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. (Color online) The baryon-strangeness correlation view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (Color online) The pressure view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. (Color online) The view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. (Color online) The trace anomaly ∆ and entropy view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. (Color online) The temperature-dependent susceptibilities view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. (Color online) Same as in fig view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. (Color online) The specific heat view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. (Color online) The pressure view at source ↗
read the original abstract

The equation of state (EoS) of strongly interacting matter at finite temperature and chemical potentials (baryon, charge, and strangeness) is a crucial input for hydrodynamic simulations of relativistic heavy-ion collisions. We construct a four-dimensional EoS using a deep-learning-assisted quasi-particle model (DLQPM) within a physics-informed neural network (PINN) framework, in which the masses of light quarks, strange quarks, and gluons are parameterized as functions of temperature and chemical potentials ($T, \mu_B, \mu_Q, \mu_S$). The model is constrained by lattice QCD data at vanishing chemical potentials and provides a thermodynamically consistent extrapolation to finite $\mu_{B,Q,S}$. The DLQPM accurately reproduces the lattice-calculated cumulants $\chi^{B,Q,S}_{i,j,k}$ at $\mu_{B,Q,S}=0$, and its predicted EoS at various chemical potentials agrees well with results from the generalized $T'$-expansion method in lattice QCD. Furthermore, the calculated baryon-strangeness correlation $C_{BS}$ is consistent, within uncertainties, with preliminary STAR data. This work offers a reliable EoS for exploring the QCD phase structure in the beam energy scan region.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces a deep-learning-assisted quasi-particle model (DLQPM) using physics-informed neural networks (PINNs) to parameterize the masses of light quarks, strange quarks, and gluons as functions of temperature T and chemical potentials μ_B, μ_Q, μ_S. The model is trained exclusively on lattice QCD data at vanishing chemical potentials to enforce thermodynamic consistency via the PINN loss and generates a four-dimensional EoS. It claims accurate reproduction of lattice cumulants χ^{B,Q,S}_{i,j,k} at μ=0, good agreement of the finite-μ EoS with the generalized T'-expansion method, and consistency (within uncertainties) of the baryon-strangeness correlation C_BS with preliminary STAR data.

Significance. If the finite-μ extrapolation proves reliable, the DLQPM EoS would supply a practical, thermodynamically consistent input for hydrodynamic simulations of heavy-ion collisions in the beam-energy-scan regime. The PINN framework's enforcement of thermodynamic relations by construction is a methodological strength that could be extended to other quasi-particle approaches.

major comments (3)
  1. [Abstract] Abstract: The central claim that the DLQPM 'accurately reproduces' the lattice cumulants χ^{B,Q,S}_{i,j,k} at μ_{B,Q,S}=0 is presented without any quantitative metrics (e.g., relative errors, χ² values, or error bands on the reproduced quantities), details on the PINN architecture, loss-function weights, or fitting procedure. This absence directly undermines assessment of the model's fidelity to the training data.
  2. [Model and results] Model construction and results: The mass functions m_q(T, μ_B, μ_Q, μ_S), m_s, and m_g are parameterized and fitted solely to μ=0 lattice pressure, energy density, and cumulants; the finite-μ EoS is then generated by the same functional form. Because no independent lattice data at finite μ are used for validation, the reported agreement with the generalized T'-expansion method compares two extrapolations rather than testing against direct data, leaving the physical correctness of the μ-dependence unverified.
  3. [Methodology] Thermodynamic consistency: While the PINN loss enforces thermodynamic relations by construction, this does not constrain the functional form of the mass parameterization itself. Consequently, thermodynamic consistency alone cannot guarantee that the extrapolated EoS at finite μ is free of systematic bias inherited from the chosen mass ansatz.
minor comments (1)
  1. [Abstract] Notation for the cumulants χ^{B,Q,S}_{i,j,k} and the correlation C_BS should be defined explicitly on first use, including the precise definitions of the indices.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough and constructive review. We address each major comment below and indicate planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that the DLQPM 'accurately reproduces' the lattice cumulants χ^{B,Q,S}_{i,j,k} at μ_{B,Q,S}=0 is presented without any quantitative metrics (e.g., relative errors, χ² values, or error bands on the reproduced quantities), details on the PINN architecture, loss-function weights, or fitting procedure. This absence directly undermines assessment of the model's fidelity to the training data.

    Authors: We agree that the abstract would be strengthened by quantitative support for the reproduction claim. In the revised manuscript we will add explicit metrics (maximum relative deviations and χ² per degree of freedom for the cumulants) and will include concise statements on the PINN architecture, loss-function weighting, and training procedure. These details already appear in Section II; we will ensure they are also summarized at the beginning of the paper for immediate accessibility. revision: yes

  2. Referee: [Model and results] Model construction and results: The mass functions m_q(T, μ_B, μ_Q, μ_S), m_s, and m_g are parameterized and fitted solely to μ=0 lattice pressure, energy density, and cumulants; the finite-μ EoS is then generated by the same functional form. Because no independent lattice data at finite μ are used for validation, the reported agreement with the generalized T'-expansion method compares two extrapolations rather than testing against direct data, leaving the physical correctness of the μ-dependence unverified.

    Authors: We acknowledge that no direct lattice data at finite chemical potentials exist for independent validation, owing to the sign problem. The comparison with the T'-expansion method is therefore between two extrapolation procedures. We will revise the text to state this limitation explicitly, to frame the agreement as a cross-check between independent methods rather than a direct test, and to quantify the spread between the two approaches as an estimate of extrapolation uncertainty. revision: partial

  3. Referee: [Methodology] Thermodynamic consistency: While the PINN loss enforces thermodynamic relations by construction, this does not constrain the functional form of the mass parameterization itself. Consequently, thermodynamic consistency alone cannot guarantee that the extrapolated EoS at finite μ is free of systematic bias inherited from the chosen mass ansatz.

    Authors: The referee correctly notes that thermodynamic consistency enforced by the loss function does not remove possible systematic bias from the quasi-particle mass ansatz. In the revised manuscript we will add an explicit discussion of this model dependence, including sensitivity tests to neural-network depth/width and alternative mass functional forms, together with a statement on the remaining extrapolation uncertainty. revision: yes

Circularity Check

0 steps flagged

No significant circularity; model fit and extrapolation are independent of target outputs

full rationale

The derivation trains a PINN-parameterized quasi-particle mass function exclusively on lattice data at μ=0 (pressure, energy density, and cumulants), with thermodynamic relations enforced by construction in the loss. The four-dimensional EoS at finite μ is then obtained by direct evaluation of the trained integrals. This is a standard extrapolation whose finite-μ values are not equivalent to the μ=0 inputs by construction, nor are they statistically forced to match any particular form; the μ-dependence remains an unconstrained output of the network architecture. No load-bearing self-citations, uniqueness theorems, or ansatz smuggling appear in the provided text, and the comparison to the T'-expansion method is an external cross-check rather than an internal reduction. The central claim therefore retains independent content beyond its training data.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on fitting effective masses of light quarks, strange quarks, and gluons as functions of T, μ_B, μ_Q, μ_S to lattice QCD data at μ=0 inside a PINN framework that enforces thermodynamic relations; extrapolation to finite μ relies entirely on the chosen functional form of those masses.

free parameters (1)
  • Parameters of the mass functions m_q(T, μ_B, μ_Q, μ_S), m_s(T, μ_B, μ_Q, μ_S), m_g(T, μ_B, μ_Q, μ_S)
    These are determined by fitting the neural network to lattice cumulants at vanishing chemical potentials.
axioms (2)
  • domain assumption Quasi-particle description remains valid at finite chemical potentials
    Core modeling assumption of the DLQPM framework.
  • domain assumption Physics-informed neural network architecture guarantees thermodynamic consistency
    Invoked to ensure the trained model satisfies thermodynamic identities.

pith-pipeline@v0.9.0 · 5525 in / 1560 out tokens · 67169 ms · 2026-05-08T09:22:11.918943+00:00 · methodology

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

68 extracted references · 52 canonical work pages · 1 internal anchor

  1. [1]

    C. Shen, Z. Qiu, H. Song, J. Bernhard, S. Bass, and U. Heinz, Comput. Phys. Commun.199, 61 (2016), arXiv:1409.8164 [nucl-th]

    work page arXiv 2016

  2. [2]

    L. Pang, Q. Wang, and X.-N. Wang, Phys. Rev. C86, 024911 (2012), arXiv:1205.5019 [nucl-th]

    work page arXiv 2012

  3. [3]

    L.-G. Pang, H. Petersen, and X.-N. Wang, Phys. Rev. C 97, 064918 (2018), arXiv:1802.04449 [nucl-th]

    work page arXiv 2018

  4. [4]

    J. E. Bernhard, J. S. Moreland, and S. A. Bass, Nature Phys.15, 1113 (2019)

    2019

  5. [5]

    Achenbachet al., Nucl

    P. Achenbachet al., Nucl. Phys. A1047, 122874 (2024), arXiv:2303.02579 [hep-ph]

    work page arXiv 2024

  6. [6]

    L.-G. Pang, K. Zhou, N. Su, H. Petersen, H. St¨ ocker, and X.-N. Wang, Nature Commun.9, 210 (2018), arXiv:1612.04262 [hep-ph]

    work page arXiv 2018

  7. [7]

    Properties of hot and dense matter from relativistic heavy ion collisions,

    P. Braun-Munzinger, V. Koch, T. Sch¨ afer, and J. Stachel, Phys. Rept.621, 76 (2016), arXiv:1510.00442 [nucl-th]

    work page arXiv 2016

  8. [8]

    Elfner and B

    H. Elfner and B. M¨ uller, J. Phys. G50, 103001 (2023), arXiv:2210.12056 [nucl-th]

    work page arXiv 2023

  9. [9]

    Monnai, B

    A. Monnai, B. Schenke, and C. Shen, Phys. Rev. C100, 024907 (2019), arXiv:1902.05095 [nucl-th]

    work page arXiv 2019

  10. [10]

    Monnai, B

    A. Monnai, B. Schenke, and C. Shen, Int. J. Mod. Phys. A36, 2130007 (2021), arXiv:2101.11591 [nucl-th]

    work page arXiv 2021

  11. [11]

    G. S. Denicol, C. Gale, S. Jeon, A. Monnai, B. Schenke, and C. Shen, Phys. Rev. C98, 034916 (2018), arXiv:1804.10557 [nucl-th]

    work page arXiv 2018

  12. [12]

    Sorensen et al., Dense nuclear matter equation of state from heavy-ion collisions, Prog

    A. Sorensenet al., Prog. Part. Nucl. Phys.134, 104080 (2024), arXiv:2301.13253 [nucl-th]

    work page arXiv 2024

  13. [13]

    Anet al., The BEST framework for the search for the QCD critical point and the chiral magnetic effect, Nucl

    X. Anet al., Nucl. Phys. A1017, 122343 (2022), arXiv:2108.13867 [nucl-th]

    work page arXiv 2022

  14. [14]

    Kumaret al.(MUSES), Living Rev

    R. Kumaret al.(MUSES), Living Rev. Rel.27, 3 (2024), arXiv:2303.17021 [nucl-th]

    work page arXiv 2024

  15. [15]

    Wu, G.-Y

    X.-Y. Wu, G.-Y. Qin, L.-G. Pang, and X.-N. Wang, Phys. Rev. C105, 034909 (2022), arXiv:2107.04949 [hep- ph]

    work page arXiv 2022

  16. [16]

    Borsanyi, Z

    S. Borsanyi, Z. Fodor, C. Hoelbling, S. D. Katz, S. Krieg, and K. K. Szabo, Phys. Lett. B730, 99 (2014), arXiv:1309.5258 [hep-lat]

    work page arXiv 2014

  17. [17]

    Bazavovet al

    A. Bazavovet al.(HotQCD), Phys. Lett. B795, 15 (2019), arXiv:1812.08235 [hep-lat]

    work page arXiv 2019

  18. [18]

    Bazavovet al.(HotQCD), Phys

    A. Bazavovet al.(HotQCD), Phys. Rev. D86, 034509 (2012), arXiv:1203.0784 [hep-lat]

    work page arXiv 2012

  19. [19]

    Bazavovet al

    A. Bazavovet al.(HotQCD), Phys. Rev. D90, 094503 (2014), arXiv:1407.6387 [hep-lat]

    work page arXiv 2014

  20. [20]

    Bollweg, J

    D. Bollweg, J. Goswami, O. Kaczmarek, F. Karsch, S. Mukherjee, P. Petreczky, C. Schmidt, and P. Scior (HotQCD), Phys. Rev. D105, 074511 (2022), 7 arXiv:2202.09184 [hep-lat]

    work page arXiv 2022

  21. [21]

    Barra and P

    D. Bollweg, J. Goswami, O. Kaczmarek, F. Karsch, S. Mukherjee, P. Petreczky, C. Schmidt, and P. Scior (HotQCD), Phys. Rev. D104(2021), 10.1103/Phys- RevD.104.074512, arXiv:2107.10011 [hep-lat]

    work page doi:10.1103/phys- 2021

  22. [22]

    Mondal, S

    S. Mondal, S. Mukherjee, and P. Hegde, Phys. Rev. Lett. 128, 022001 (2022), arXiv:2106.03165 [hep-lat]

    work page arXiv 2022

  23. [23]

    Bors´ anyi, Z

    S. Bors´ anyi, Z. Fodor, J. N. Guenther, R. Kara, S. D. Katz, P. Parotto, A. P´ asztor, C. Ratti, and K. K. Szab´ o, Phys. Rev. Lett.126, 232001 (2021), arXiv:2102.06660 [hep-lat]

    work page arXiv 2021

  24. [24]

    Borsanyi, J

    S. Borsanyi, J. N. Guenther, R. Kara, Z. Fodor, P. Parotto, A. Pasztor, C. Ratti, and K. K. Szabo, Phys. Rev. D105, 114504 (2022), arXiv:2202.05574 [hep-lat]

    work page arXiv 2022

  25. [25]

    Monnai, G

    A. Monnai, G. Pihan, B. Schenke, and C. Shen, Phys. Rev. C110, 044905 (2024), arXiv:2406.11610 [nucl-th]

    work page arXiv 2024

  26. [26]

    Abuali, S

    A. Abuali, S. Bors´ anyi, Z. Fodor, J. Jahan, M. Ka- hangirwe, P. Parotto, A. P´ asztor, C. Ratti, H. Shah, and S. A. Trabulsi, Phys. Rev. D112, 054502 (2025), arXiv:2504.01881 [hep-lat]

    work page arXiv 2025

  27. [27]

    Soloveva, A

    O. Soloveva, A. Palermo, and E. Bratkovskaya, Phys. Rev. C110, 034908 (2024), arXiv:2311.15984 [hep-ph]

    work page arXiv 2024

  28. [28]

    Z. Kang, J. Zhu, and J. Guo, Phys. Rev. D107, 076005 (2023), arXiv:2211.09442 [hep-ph]

    work page arXiv 2023

  29. [29]

    Peshier, B

    A. Peshier, B. Kampfer, O. P. Pavlenko, and G. Soff, Phys. Rev. D54, 2399 (1996)

    1996

  30. [30]

    Alqahtani, M

    M. Alqahtani, M. Nopoush, and M. Strickland, Phys. Rev. C92, 054910 (2015), arXiv:1509.02913 [hep-ph]

    work page arXiv 2015

  31. [31]

    Liu, W.-J

    F.-L. Liu, W.-J. Xing, X.-Y. Wu, G.-Y. Qin, S. Cao, and X.-N. Wang, Eur. Phys. J. C82, 350 (2022), arXiv:2107.11713 [hep-ph]

    work page arXiv 2022

  32. [32]

    Soloveva, J

    O. Soloveva, J. Aichelin, and E. Bratkovskaya, Phys. Rev. D105, 054011 (2022), arXiv:2108.08561 [hep-ph]

    work page arXiv 2022

  33. [33]

    Plumari, W

    S. Plumari, W. M. Alberico, V. Greco, and C. Ratti, Phys. Rev. D84, 094004 (2011), arXiv:1103.5611 [hep- ph]

    work page arXiv 2011

  34. [34]

    Li, H.-L

    F.-P. Li, H.-L. L¨ u, L.-G. Pang, and G.-Y. Qin, Phys. Lett. B844, 138088 (2023), arXiv:2211.07994 [hep-ph]

    work page arXiv 2023

  35. [35]

    Li, L.-G

    F.-P. Li, L.-G. Pang, and G.-Y. Qin, Phys. Lett. B868, 139692 (2025), arXiv:2501.10012 [nucl-th]

    work page arXiv 2025

  36. [36]

    Mykhaylova and C

    V. Mykhaylova and C. Sasaki, Phys. Rev. D103, 014007 (2021), arXiv:2007.06846 [hep-ph]

    work page arXiv 2021

  37. [37]

    Romatschke, Phys

    P. Romatschke, Phys. Rev. D85, 065012 (2012), arXiv:1108.5561 [gr-qc]

    work page arXiv 2012

  38. [38]

    Mykhaylova,Transport Properties of Hot QCD Matter in the Quasiparticle Approach, Ph.D

    V. Mykhaylova,Transport Properties of Hot QCD Matter in the Quasiparticle Approach, Ph.D. thesis

  39. [39]

    Thuerey, P

    N. Thuerey, P. Holl, M. Mueller, P. Schnell, F. Trost, and K. Um,Physics-based Deep Learning(WWW, 2021)

    2021

  40. [40]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Rai- son, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, inAdvances in Neural Informa- tion Processing Systems, Vol. 32, edited by H. Wallach, H. Larochelle, A. Beygelzimer,...

    2019

  41. [41]

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, Journal of Machine Learning Research18, 1 (2018)

    2018

  42. [42]

    Hornik, M

    K. Hornik, M. B. Stinchcombe, and H. L. White, Neural Networks2, 359 (1989)

    1989

  43. [43]

    Mattuck, Osborne McGraw-Hill (1976)

    R. Mattuck, Osborne McGraw-Hill (1976)

    1976

  44. [44]

    J. I. Kapusta and C. Gale,Finite-temperature field theory: Principles and applications, Cambridge Mono- graphs on Mathematical Physics (Cambridge University Press, 2011)

    2011

  45. [45]

    P. P. M. Raissi and G. Karniadakis, Journal of Compu- tational Physics378, 686 (2019)

    2019

  46. [46]

    Raissi, P

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

    2019

  47. [47]

    L. Wang, J. Subatomic Part. Cosmol.3, 100024 (2025), arXiv:2501.00374 [nucl-th]

    work page arXiv 2025

  48. [48]

    S. Soma, L. Wang, S. Shi, H. St¨ ocker, and K. Zhou, EPJ Web Conf.276, 06007 (2023)

    2023

  49. [49]

    Pang, Int

    L.-G. Pang, Int. J. Mod. Phys. E33, 2430009 (2024)

    2024

  50. [50]

    S. Shi, K. Zhou, J. Zhao, S. Mukherjee, and P. Zhuang, Phys. Rev. D105, 014017, arXiv:2105.07862 [hep-ph]

    work page arXiv

  51. [51]

    K. Zhou, L. Wang, L.-G. Pang, and S. Shi, Prog. Part. Nucl. Phys.135, 104084 (2024), arXiv:2303.15136 [hep- ph]

    work page arXiv 2024

  52. [52]

    Aarts, K

    G. Aarts, K. Fukushima, T. Hatsuda, A. Ipp, S. Shi, L. Wang, and K. Zhou, Nature Reviews Physics7, 1 (2025)

    2025

  53. [53]

    V. Koch, A. Majumder, and J. Randrup, Phys. Rev. Lett.95, 182301 (2005), arXiv:nucl-th/0505052

    work page arXiv 2005

  54. [54]

    V. Koch, A. Majumder, and J. Randrup, Nucl. Phys. A 774, 841 (2006)

    2006

  55. [55]

    Baryon- Strangeness Correlations in Au+Au Collisions at RHIC-STAR, XXXI International Conference on Ultra- relativistic Nucleus-Nucleus Collisions, Frankfurt, Apr.6- 12, 2025,

    H. Feng (for the STAR Collaboration), “Baryon- Strangeness Correlations in Au+Au Collisions at RHIC-STAR, XXXI International Conference on Ultra- relativistic Nucleus-Nucleus Collisions, Frankfurt, Apr.6- 12, 2025,”

    2025

  56. [56]

    Luo and N

    X. Luo and N. Xu, Nucl. Sci. Tech.28, 112 (2017), arXiv:1701.02105 [nucl-ex]

    work page arXiv 2017

  57. [57]

    Chenget al., Phys

    M. Chenget al., Phys. Rev. D79, 074505 (2009), arXiv:0811.1006 [hep-lat]

    work page arXiv 2009

  58. [58]

    Bazavovet al., Phys

    A. Bazavovet al., Phys. Rev. Lett.113, 072001 (2014), arXiv:1404.6511 [hep-lat]

    work page arXiv 2014

  59. [59]

    Bazavovet al., Phys

    A. Bazavovet al., Phys. Rev. Lett.111, 082301 (2013), arXiv:1304.7220 [hep-lat]

    work page arXiv 2013

  60. [60]

    Bollweg, H

    D. Bollweg, H. T. Ding, J. Goswami, F. Karsch, S. Mukherjee, P. Petreczky, and C. Schmidt, Phys. Rev. D110, 054519 (2024), arXiv:2407.09335 [hep-lat]

    work page arXiv 2024

  61. [61]

    Hadronic Fluctuations and Correlations,

    V. Koch, “Hadronic Fluctuations and Correlations,” in Relativistic Heavy Ion Physics, edited by R. Stock (2010) pp. 626–652, arXiv:0810.2520 [nucl-th]

    work page arXiv 2010

  62. [62]

    Koch and V

    V. Koch and V. Vovchenko, (2025), arXiv:2512.04288 [nucl-th]

    work page arXiv 2025

  63. [63]

    Searching for Activation Functions

    P. Ramachandran, B. Zoph, and Q. V. Le, “Searching for activation functions,” (2017), arXiv:1710.05941 [cs.NE]

    work page internal anchor Pith review arXiv 2017

  64. [64]

    Haque and M

    N. Haque and M. G. Mustafa, Prog. Part. Nucl. Phys. 140, 104136 (2025), arXiv:2404.08734 [hep-ph]

    work page arXiv 2025

  65. [65]

    Levai and U

    P. Levai and U. W. Heinz, Phys. Rev. C57, 1879 (1998), arXiv:hep-ph/9710463

    work page arXiv 1998

  66. [66]

    Adamczyket al.(STAR), Phys

    L. Adamczyket al.(STAR), Phys. Rev. C96, 044904 (2017), arXiv:1701.07065 [nucl-ex]

    work page arXiv 2017

  67. [67]

    H. T. Ding, S. T. Li, Q. Shi, and X. D. Wang, Eur. Phys. J. A57, 202 (2021), arXiv:2104.06843 [hep-lat]

    work page arXiv 2021

  68. [68]

    Kumar, H.-T

    A. Kumar, H.-T. Ding, J.-B. Gu, and S.-T. Li, PoS LA TTICE2024, 175 (2025), arXiv:2502.03152 [hep-lat]. Appendix A: The training results Fig. 5 illustrates the temperature dependence of both the trace anomaly ∆ and the entropy densitys/T 3 at µB = 0. The results indicate that the DLQPM can ac- 8 curately reproduce the QCD EoS. As temperature in- creases, ...

    work page arXiv 2025