pith. sign in

arxiv: 2506.03837 · v2 · pith:FICQDBLZnew · submitted 2025-06-04 · ❄️ cond-mat.supr-con · cond-mat.mtrl-sci· cs.AI· cs.LG

HTSC-2025: A Benchmark Dataset of Ambient-Pressure High-Temperature Superconductors for AI-Driven Critical Temperature Prediction

Xiao-Qi Han , Ze-Feng Gao , Xin-De Wang , Zhenfeng Ouyang , Peng-Jie Guo , Zhong-Yi Lu This is my paper

Pith reviewed 2026-05-22 00:44 UTC · model grok-4.3

classification ❄️ cond-mat.supr-con cond-mat.mtrl-scics.AIcs.LG
keywords high-temperature superconductorsambient-pressure superconductorsbenchmark datasetcritical temperature predictionAI machine learningBCS superconductivityhydride materials
0
0 comments X

The pith

The HTSC-2025 dataset compiles recent theoretical predictions of ambient-pressure high-temperature superconductors to serve as a standard benchmark for AI-based critical temperature prediction.

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

This paper introduces the HTSC-2025 benchmark dataset to address the absence of widely accepted standards for evaluating AI algorithms that predict superconducting critical temperatures. The dataset compiles theoretically predicted ambient-pressure high-temperature superconductors from 2023 to 2025 based on BCS theory, encompassing systems such as X2YH6, MXH3 perovskites, M3XH8, BCN-doped cage structures, and 2D honeycomb lattices. Establishing this resource enables consistent comparisons across different AI methods and supports their refinement for material discovery applications. The open-source availability and plan for continuous updates position it as a living resource for the community.

Core claim

The paper presents HTSC-2025 as a comprehensive benchmark dataset of ambient-pressure high-temperature superconductors that were theoretically predicted between 2023 and 2025 using BCS superconductivity theory. It incorporates the X₂YH₆ system, the perovskite MXH₃ system, the M₃XH₈ system, cage-like BCN-doped metal atomic systems evolved from LaH₁₀, and two-dimensional honeycomb-structured systems evolved from MgB₂. The dataset is open-sourced with a commitment to ongoing updates to promote the use of AI in identifying new superconducting materials.

What carries the argument

The HTSC-2025 benchmark dataset as a compiled collection of recent theoretical predictions for standardized AI evaluation.

Load-bearing premise

The theoretical predictions of critical temperatures and the selection of included materials are accurate and representative enough to form a trustworthy basis for benchmarking AI predictions.

What would settle it

Experimental measurement of the critical temperatures for a subset of the materials in the HTSC-2025 dataset and comparison against the theoretical values used in the benchmark would directly test its suitability.

Figures

Figures reproduced from arXiv: 2506.03837 by Peng-Jie Guo, Xiao-Qi Han, Xin-De Wang, Ze-Feng Gao, Zhenfeng Ouyang, Zhong-Yi Lu.

Figure 1
Figure 1. Figure 1: FIG. 1. Construction and Application of the HTSC-2025 Benchmark. (a) Workflow of HTSC-2025 benchmark dataset con [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Statistics and elemental composition of the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
read the original abstract

The discovery of high-temperature superconducting materials holds great significance for human industry and daily life. In recent years, research on predicting superconducting transition temperatures using artificial intelligence~(AI) has gained popularity, with most of these tools claiming to achieve remarkable accuracy. However, the lack of widely accepted benchmark datasets in this field has severely hindered fair comparisons between different AI algorithms and impeded further advancement of these methods. In this work, we present the HTSC-2025, an ambient-pressure high-temperature superconducting benchmark dataset. This comprehensive compilation encompasses theoretically predicted superconducting materials discovered by theoretical physicists from 2023 to 2025 based on BCS superconductivity theory, including the renowned X$_2$YH$_6$ system, perovskite MXH$_3$ system, M$_3$XH$_8$ system, cage-like BCN-doped metal atomic systems derived from LaH$_{10}$ structural evolution, and two-dimensional honeycomb-structured systems evolving from MgB$_2$. The HTSC-2025 benchmark has been open-sourced at https://github.com/xqh19970407/HTSC-2025 and will be continuously updated. This benchmark holds significant importance for accelerating the discovery of superconducting materials using AI-based methods.

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

2 major / 2 minor

Summary. The manuscript presents HTSC-2025, a benchmark dataset compiling theoretically predicted ambient-pressure high-temperature superconductors discovered from 2023 to 2025 based on BCS theory. It includes the X₂YH₆ system, perovskite MXH₃ system, M₃XH₈ system, cage-like BCN-doped metal atomic systems, and two-dimensional honeycomb-structured systems. The dataset is open-sourced on GitHub with plans for continuous updates and is intended to enable fair comparisons and accelerate AI-driven Tc prediction for material discovery.

Significance. If the compiled theoretical predictions prove reliable, the dataset could standardize evaluations of AI models for superconducting Tc prediction and support reproducible research in the field. The open-source release and commitment to ongoing updates represent concrete strengths that enhance accessibility and long-term utility.

major comments (2)
  1. The abstract and manuscript provide no information on curation details such as the total number of candidates screened, exclusion criteria, computed Tc uncertainties, or any cross-validation against experimental data. This absence directly affects the load-bearing claim that HTSC-2025 constitutes a trustworthy benchmark for AI models targeting real ambient-pressure superconductivity.
  2. The central utility claim rests on the assumption that BCS-based theoretical Tc values from the listed structural families (X₂YH₆, MXH₃, M₃XH₈, BCN-doped cages, 2D honeycombs) are sufficiently accurate proxies. No section addresses known limitations of DFT/Eliashberg methods for stability or Tc at P=0, leaving the benchmark's relevance to experimental discovery untested.
minor comments (2)
  1. Chemical formulas in the abstract (e.g., X2YH6, LaH10) would benefit from consistent subscript formatting and a dedicated table listing all included compounds with their reported Tc ranges.
  2. The GitHub link is given but no summary statistics (number of entries, distribution of Tc values, or coverage across structure types) appear in the text; adding these would improve immediate usability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment point by point below. Revisions have been made to improve clarity on curation and to discuss methodological limitations, while maintaining the manuscript's focus on a theoretical benchmark dataset.

read point-by-point responses

  1. Referee: The abstract and manuscript provide no information on curation details such as the total number of candidates screened, exclusion criteria, computed Tc uncertainties, or any cross-validation against experimental data. This absence directly affects the load-bearing claim that HTSC-2025 constitutes a trustworthy benchmark for AI models targeting real ambient-pressure superconductivity.

    Authors: We agree that explicit curation details strengthen the presentation of the benchmark. The HTSC-2025 dataset aggregates all relevant published theoretical predictions of ambient-pressure high-Tc superconductors from 2023–2025 that satisfy BCS-based criteria. In the revised manuscript we have added a dedicated Data Curation section describing the literature search protocol, the specific studies included for each structural family (X₂YH₆, MXH₃, M₃XH₈, BCN-doped cages, and 2D honeycombs), and the Tc values together with any uncertainties reported in the original works. We have also revised the abstract and introduction to state clearly that cross-validation against experiment is not currently possible because these entries are theoretical predictions without experimental realizations to date. The benchmark is therefore positioned as a standardized collection for AI models operating on theoretical Tc data, with the long-term aim of supporting experimental discovery. revision: yes

  2. Referee: The central utility claim rests on the assumption that BCS-based theoretical Tc values from the listed structural families (X₂YH₆, MXH₃, M₃XH₈, BCN-doped cages, 2D honeycombs) are sufficiently accurate proxies. No section addresses known limitations of DFT/Eliashberg methods for stability or Tc at P=0, leaving the benchmark's relevance to experimental discovery untested.

    Authors: We acknowledge the known limitations of DFT and Eliashberg calculations for ambient-pressure stability and Tc predictions. We have added a new subsection titled “Methodological Context and Limitations” that summarizes the principal approximations (exchange-correlation functional dependence, phonon softening issues, and the BCS/Eliashberg framework) and notes that dynamical stability at P=0 remains a theoretical prediction rather than an experimental guarantee. At the same time, we maintain that a consistent, publicly documented collection of recent theoretical candidates still provides a useful and reproducible benchmark for comparing AI models. The revised discussion now explicitly frames the dataset’s utility as enabling fair algorithmic comparisons on the current theoretical landscape, while underscoring that any link to experimental discovery is indirect and will require future experimental validation. revision: yes

Circularity Check

0 steps flagged

No circularity: dataset compilation aggregates external theoretical predictions without self-referential derivation

full rationale

The paper compiles and releases a benchmark dataset of ambient-pressure high-Tc material candidates drawn from 2023-2025 BCS-theory literature (X2YH6, MXH3, M3XH8, BCN-doped cages, 2D honeycomb systems). No equations, fitted parameters, or predictions are generated inside the paper; the contribution is aggregation and open-sourcing of pre-existing external results. Consequently there are no load-bearing steps that reduce by construction to the paper's own inputs, self-citations, or ansatzes. The work is self-contained as a data resource and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests primarily on the domain assumption that BCS-theory predictions from 2023-2025 are suitable for benchmarking; no free parameters or invented entities are introduced by the paper itself.

axioms (1)
  • domain assumption BCS superconductivity theory yields reliable candidate materials for ambient-pressure high-Tc superconductors.
    The dataset is explicitly compiled from materials discovered using BCS theory as described in the abstract.

pith-pipeline@v0.9.0 · 5784 in / 1282 out tokens · 46337 ms · 2026-05-22T00:44:59.111692+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 2 internal anchors

  1. [1]

    Renmin University of China, Beijing 100872, China

    School of Physics and Beijing Key Laboratory of Opto-electronic Functional Materials & Micro-nano Devices. Renmin University of China, Beijing 100872, China

    work page

  2. [2]

    Key Laboratory of Quantum State Construction and Manipulation (Ministry of Education), Renmin University of China, Beijing 100872, China and

    work page

  3. [3]

    HTSC-2025: A Benchmark Dataset of Ambient-Pressure High-Temperature Superconductors for AI-Driven Critical Temperature Prediction

    Hefei National Laboratory, Hefei 230088, China (Dated: September 3, 2025) The discovery of high-temperature superconducting materials holds great significance for human industry and daily life. In recent years, research on predicting superconducting transition tem- peratures using artificial intelligence (AI) has gained popularity, with most of these tool...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Onnes H K 1911 Commun Theor Phys 120

    work page 1911

  5. [5]

    arXiv https://arxiv.org/

    work page

  6. [6]

    Coombs T A, Wang Q, Shah A, Hu J, Hao L, Patel I, Wei H, Wu Y, Coombs T, Wang W 2024 Nature Reviews Electrical Engineering 1 788

    work page 2024

  7. [7]

    Yao C and Ma Y 2021 iScience 24 102541

    work page 2021

  8. [8]

    2019 Nature 574 72–75

    Pustogow A, Luo Y, Chronister A et al. 2019 Nature 574 72–75

    work page 2019

  9. [9]

    Fusion 58 103001

    Bruzzone P, Fietz W H, Minervini J V, Novikov M, Yanagi N, Zhai Y and Zheng J 2018 Nucl. Fusion 58 103001

    work page 2018

  10. [10]

    Han X Q, Wang X D, Xu M Y, Feng Z, Yao B W, Guo P J, Gao Z F and Lu Z Y 2025 Chinese Phys. Lett. 42 027403

    work page 2025

  11. [11]

    Choudhary K, Garrity K 2022 NPJ Comput. Mater. 8 244

    work page 2022

  12. [12]

    Choudhary K, DeCost B 2021 NPJ Comput. Mater. 7 185

    work page 2021

  13. [13]

    Cerqueira T F T, Fang Y W, Errea I, Sanna A and Marques M A L 2024 Adv. Funct. Mater. 34 2404043

    work page 2024

  14. [14]

    Han X Q, Ouyang Z F, Guo P J, Sun H, Gao Z F and Lu Z Y 2024 Chin. Phys. Lett. 42 047301

    work page 2024

  15. [15]

    Ouyang Z F, Yao B W, Han X Q, Guo P J, Gao Z F and Lu Z Y 2025 Phys. Rev. B 111 L140501

    work page 2025

  16. [16]

    Gibson J B, Hire A C, Dee P M, Barrera O, Geisler B, Hirschfeld P J and Hennig R G 2025 npj Comput. Mater. 11 7

    work page 2025

  17. [17]

    Li J, Fang W Q, Jin S J, Zhang T D, Wu Y L, Xu X D, Liu Y and Yao D X 2024 arXiv:2409.07721

    work page arXiv 2024

  18. [18]

    Han X Q, Guo P J, Gao Z F, Sun H and Lu Z Y 2025 arXiv:2505.09203

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  19. [19]

    Abramson J, Adler J and Dunger J 2024 Nature 630 493–500

    work page 2024

  20. [20]

    Wang V, Xu N, Liu J C, Tang G and Geng W T 2021Comput. Phys. Commun. 267 108033

    work page

  21. [21]

    Quantum ESPRESSO https://www.quantum-espresso.org/

    work page

  22. [22]

    Dolui K, Conway L J, Heil C, Strobel T A, Prasankumar R P and Pickard C J 2024 Phys. Rev. Lett. 132 166001

    work page 2024

  23. [23]

    Li B, Fan Y X, Zhai J J, Cao Z S, Zhu C, Cheng J, Liu S L and Shi Z X 2025 J. Phys. Chem. C 129 7098–7106

    work page 2025

  24. [24]

    Liu S M, Shi J J, He Y, Tian C, Zhu Y H, Wang X Q and Zhong H X 2024 Adv. Funct. Mater. 34 2315386

    work page 2024

  25. [25]

    Tian C, He Y, Zhu Y H, Du J, Liu S M, Guo W H, Zhong H X, Lu J, Wang X Q and Shi J J 2024 Adv. Funct. Mater. 34 2304919

    work page 2024

  26. [26]

    He Y, Lu J, Wang X Q and Shi J J 2023 Phys. Rev. B 108 054515

    work page 2023

  27. [27]

    Cerqueira T F T, Sanna A and Marques M A L 2024 Adv. Mater. 36 2307085

    work page 2024

  28. [28]

    Futures 3 025602

    Wines D and Choudhary K 2024 Mater. Futures 3 025602

    work page 2024

  29. [29]

    Li B, Zhu C, Zhai J J, Yin C H, Fan Y X, Cheng J, Liu S L and Shi Z X 2024 Phys. Rev. B 110 214504

    work page 2024

  30. [30]

    Quan H, Shi X B, Han Y L, Zhang P and Wang B T 2025 Phys. Rev. B 111 134509

    work page 2025

  31. [31]

    Wang H, Tse J S, Tanaka K, Iitaka T and Ma Y M 2012 Proc. Natl. Acad. Sci. USA 109 6463

    work page 2012

  32. [32]

    Ma L, Wang K, Xie Y, Yang X, Wang Y Y, Zhou M, Liu H Y, Yu X H, Zhao Y S, Wang H B, Liu G T and Ma Y M 2022 Phys. Rev. Lett. 128 167001

    work page 2022

  33. [33]

    Li Y W, Hao J, Liu H Y, Tse J S, Wang Y C and Ma Y M 2015 Sci. Rep. 5 9948

    work page 2015

  34. [34]

    Troyan I A, Semenok D V, Kvashnin A G, Sadakov A V, Sobolevskiy O A, Pudalov V M and Oganov A R 2021 Adv. Mater. 33 2006832

    work page 2021

  35. [35]

    Chen W H, Semenok D V, Huang X L, Shu H Y, Li X, Duan D F, Cui T and Oganov A R 2021 Phys. Rev. Lett. 127 117001

    work page 2021

  36. [36]

    Somayazulu M, Ahart M, Mishra A K, Geballe Z M, Baldini M, Meng Y, Struzhkin V V and Hemley R J 2019 Phys. Rev. Lett. 122 027001

    work page 2019

  37. [37]

    Drozdov A P, Kong P P, Minkov V S, Besedin S P, Ku- zovnikov M A, Mozaffari S, Balicas L, Balakirev F F, Graf D E, Prakapenka V B, Greenberg E, Knyazev D A, Tkacz M and Eremets M I 2019 Nature 569 528

    work page 2019

  38. [38]

    Li B, Fan Y X, Yin C H, Zhai J J, Zhu C, Cao Z S, Cheng J and Liu S L 2025 Phys. Rev. B 111 014510 7

    work page 2025

  39. [39]

    Ding H B, Feng Y J, Jiang M J, Tian H L, Zhong G H, Yang C L, Chen X J and Lin H Q 2022 Phys. Rev. B 106 104508

    work page 2022

  40. [40]

    Li X, Yong X, Wu M, Lu S Y, Liu H Y, Meng S, Tse J S and Li Y W 2019 J. Phys. Chem. Lett. 10 2554–2560

    work page 2019

  41. [41]

    Duan Q Z, Zhan L H, Shen J Y, Zhong X and Lu C 2024 Phys. Rev. B 109 054505

    work page 2024

  42. [42]

    Geng N, Hilleke K P, Zhu L, Wang X Y, Strobel T A and Zurek E 2023 J. Am. Chem. Soc. 145 1696–1706

    work page 2023

  43. [43]

    Chen C, Miao J Y, Zuo J N, He S and Lu C 2024 Phys. Rev. B 110 174514

    work page 2024

  44. [44]

    Jiang Q W, Chen L, Ma H, Li C D, Duan D F and Cui T 2025 Inorg. Chem. 64 [DOI: 10.1021/acs.inorgchem.4c04932]

    work page doi:10.1021/acs.inorgchem.4c04932 2025

  45. [45]

    Di Cataldo S and Boeri L 2023 Phys. Rev. B 107 L060501

    work page 2023

  46. [46]

    Han Y L, Jiang K Y, Wang B T, Zhang P and Lu H Y 2025 Phys. Rev. B 111 094520

    work page 2025

  47. [47]

    Tomassetti C R, Gochitashvili D, Renskers C, Margine E R and Kolmogorov A N 2024 Phys. Rev. Mater. 8 114801

    work page 2024

  48. [48]

    Liu H D, Fu X P, Fu Z G, Lu H Y and Zhang P 2025 Phys. Rev. B 111 184502

    work page 2025

  49. [49]

    Nagamatsu J, Nakagawa N, Muranaka T, Zenitani Y and Akimitsu J 2001 Nature 410 63

    work page 2001

  50. [50]

    Singh S, Romero A H, Mella J D, Eremeev V, Mu˜ noz E, Alexandrova A N, Rabe K M, Vanderbilt D and Mu˜ noz F 2022 npj Quantum Mater. 7 37

    work page 2022

  51. [51]

    Jiang J J, Xue Y M, Zha L L, Yao S W, Wang B, Hu W J, Peng L, Shi T T, Chen J, Liu X L and Lin J 2025 J. Mater. Chem. C

    work page 2025

  52. [52]

    Liu W Y, Wang X J, Li Y Q, Wei Y H, Zhong M M and Kuang M Q 2025 Supercond. Sci. Technol. 38 045024

    work page 2025

  53. [53]

    Jiang B, Luo X S, Sun Y, Zhong X, Lv J, Xie Y, Ma Y M and Liu H Y 2025 Phys. Rev. B 111 054505

    work page 2025

  54. [54]

    Wang X Y, Zhang C Q, Wang Z Y, Liu H Y, Lv J, Wang H, E W N and Ma Y M 2025 arXiv:2502.16558

    work page arXiv 2025