Reinterpretation and preservation of data and analyses in HEP

J. M. Butterworth, Sabine Kraml, H. Prosper, A. G. Buckley, Louie Corpe, C. Diaconu, Mark Goodsell, P. Gras, Martin Habedank, C. Lange, K. Lassila-Perini, André Lessa, Rakhi Mahbubani, J. Mamuzic, Zach Marshall, T. McCauley, Humberto Reyes-González, Krzysztof Rolbiecki, S. Sekmen, G. H. Stark, G. Watt, Jonas Würzinger, Shehu AbdusSalam, Aytul Adiguzel, Amine Ahriche, B. C. Allanach, Mohammad Mahdi Altakach, Jack Y. Araz, Etera R. Livine, Saiyad Ashanujjaman, V. A. Austrup, Emanuele Bagnaschi, Sumit Banik, Csaba Balázs, Daniele Barducci, P. Bechtle, Samuel Bein, N. Berger, Tonmoy Biswas, F. Boudjema, Jamie Boyd, Carsten Burgard, J. C. Burzynski, Jordan Byers, Giacomo Cacciapaglia, Cécile Caillol, Orhan Çakir, Christopher Chang, A. Chen, A. Coccaro, Yara Do Amaral Coutinho, Andreas Crivellin, Lucas Constantin, G. Cottin, Hridoy Debnath, M. Demirci, Juhi Dutta, J. Franklin Egan, Carlos Erice Cid, F. Fassi, M. Feickert, Arnaud Ferrari, P. Perez, Des Fitzgerald, Roberto Franceschini, Benjamin Fuks, Lorenz Gärtner, Kirtiman Ghosh, A. Giammanco, T. A. Gómez Espinosa, Leticia Guedes, Giovanni Guerrieri, C. Gütschow, Abdelhamid Haddad, Mahsana Haleem, H. Hamdaoui, S. Heinemeyer, L. Heinrich, Ben Hodkinson, Gabriela Hoff, Cyril Hugonie, S. Jeon, Adil Jueid, D. Kar, A. Kaczmarská, Venus Keus, Michael Klasen, Kyoungchul Kong, Joachim Kopp, Michael Krämer, Manuel Kunkel, Bertrand Laforge, Theodota Lagouri, E. Lançon, Peilian Li, Gabriela Lichtenstein, Y. Liu, S. Lowette, Jayita Lahiri, Siddharth P. MaharathyarXiv 2025 · DOI · arXiv

Abstract: Data from particle physics experiments are unique and are often the result of a very large investment of resources. Given the potential scientific impact of these data, which goes far beyond the immediate priorities of the experimental collaborations that obtain them, it is imperative that the collaborations and the wider particle physics community publish and preserve sufficient information to ensure that this impact can be realised, now and into the future. The information to be published and preserved includes the algorithms, statistical information, simulations and the recorded data. This publication and preservation requires significant resources, and should be a strategic priority with commensurate planning and resource allocation from the earliest stages of future facilities and experiments.

Introduction: The Challenge of Data Longevity

In high-energy physics (HEP), experimental results are typically published to test specific theoretical models. However, as theoretical physics evolves, scientists often need to "re-interpret" existing data to evaluate new hypotheses. Currently, the field faces a significant "shelf-life" problem: the specialized knowledge required to reproduce or re-run an analysis often disappears when the original researchers move on.

Evidence suggests that the reproducibility of several LHC Run-1 and Run-2 searches has already been lost even within the collaborations that created them. Without standardized preservation of analysis logic, statistical models, and detector simulations, the massive global investment in experiments like the LHC risks becoming a "write-only" archive. This paper from the LHC Reinterpretation Forum issues a "call to action" to treat data preservation as a strategic priority.

A Strategic Roadmap for Preservation

To address these challenges, the paper outlines a multi-layered approach to data and analysis preservation based on the "FAIR" principles (Findable, Accessible, Interoperable, and Reusable). Key methodologies include:

  • Containerization: Utilizing tools like REANA to encapsulate software environments, ensuring code remains executable across different operating systems and over long periods.
  • Standardized Statistical Models: Moving from static plots to digital likelihoods using tools like pyhf and the HistFactory standard, allowing for precise re-evaluation of results.
  • Surrogate Modeling: Implementing Machine Learning (ML) to create models of complex detectors, enabling theorists to simulate detector responses without requiring proprietary experiment software.
  • Analysis Description Languages (ADL): Developing human-readable, framework-independent languages to describe analysis logic.

These efforts are supported by an ecosystem of data sharing, including the CERN Open Data Portal (hosting petabytes of collision data) and HEPData, the standard repository for digitized tables and statistical likelihoods.

The Impact of Open Science

One of the most compelling arguments for open science in HEP is the quantitative benefit to researchers. Publications that provide a HEPData entry or a re-implementation of their analysis logic see a significant increase in citations—averaging 22% higher overall, and up to 90% higher for searches involving physics beyond the Standard Model (BSM).

Growth of third-party publications using CMS Open Data over time, demonstrating the scientific value of accessible datasets.

As shown in Figure 1, the availability of CMS Open Data has led to a steady explosion of third-party research papers, proving that external researchers can produce high-quality science when provided with the necessary tools and data. However, adoption is not yet universal across the community.

Comparison of total LHC publications versus those with associated HEPData records across major collaborations, highlighting the remaining gap in data sharing.

Figure 2 illustrates a disparity between major collaborations. While ATLAS and CMS have high rates of data sharing, a significant portion of LHC publications still lack associated digital records on HEPData, representing a missed opportunity for future reinterpretation.

Future Directions and Resource Allocation

Preservation is not a cost-free endeavor; it requires significant human effort and financial investment from the earliest stages of an experiment's design. The authors argue for a shift in career incentives, where the "physicist-programmer" responsible for data preservation receives professional credit comparable to those making primary discoveries.

Furthermore, future data releases must address the complexity of "long-lived particles" and other specialized use cases by providing more "raw" data formats. By making reinterpretation a strategic priority, the HEP community can ensure that the scientific value of its unique datasets is maximized for decades to come.

Summary updated

Abstract

Data from particle physics experiments are unique and are often the result of a very large investment of resources. Given the potential scientific impact of these data, which goes far beyond the immediate priorities of the experimental collaborations that obtain them, it is imperative that the collaborations and the wider particle physics community publish and preserve sufficient information to ensure that this impact can be realised, now and into the future. The information to be published and preserved includes the algorithms, statistical information, simulations and the recorded data. This publication and preservation requires significant resources, and should be a strategic priority with commensurate planning and resource allocation from the earliest stages of future facilities and experiments.
BibTeX
@misc{butterworth2025reinterpretation,
  author = {J. M. Butterworth and Sabine Kraml and H. Prosper and A. G. Buckley and Louie Corpe and C. Diaconu and Mark Goodsell and P. Gras and Martin Habedank and C. Lange and K. Lassila-Perini and André Lessa and Rakhi Mahbubani and J. Mamuzic and Zach Marshall and T. McCauley and Humberto Reyes-González and Krzysztof Rolbiecki and S. Sekmen and G. H. Stark and G. Watt and Jonas Würzinger and Shehu AbdusSalam and Aytul Adiguzel and Amine Ahriche and B. C. Allanach and Mohammad Mahdi Altakach and Jack Y. Araz and Etera R. Livine and Saiyad Ashanujjaman and V. A. Austrup and Emanuele Bagnaschi and Sumit Banik and Csaba Balázs and Daniele Barducci and P. Bechtle and Samuel Bein and N. Berger and Tonmoy Biswas and F. Boudjema and Jamie Boyd and Carsten Burgard and J. C. Burzynski and Jordan Byers and Giacomo Cacciapaglia and Cécile Caillol and Orhan Çakir and Christopher Chang and A. Chen and A. Coccaro and Yara Do Amaral Coutinho and Andreas Crivellin and Lucas Constantin and G. Cottin and Hridoy Debnath and M. Demirci and Juhi Dutta and J. Franklin Egan and Carlos Erice Cid and F. Fassi and M. Feickert and Arnaud Ferrari and P. Perez and Des Fitzgerald and Roberto Franceschini and Benjamin Fuks and Lorenz Gärtner and Kirtiman Ghosh and A. Giammanco and T. A. Gómez Espinosa and Leticia Guedes and Giovanni Guerrieri and C. Gütschow and Abdelhamid Haddad and Mahsana Haleem and H. Hamdaoui and S. Heinemeyer and L. Heinrich and Ben Hodkinson and Gabriela Hoff and Cyril Hugonie and S. Jeon and Adil Jueid and D. Kar and A. Kaczmarská and Venus Keus and Michael Klasen and Kyoungchul Kong and Joachim Kopp and Michael Krämer and Manuel Kunkel and Bertrand Laforge and Theodota Lagouri and E. Lançon and Peilian Li and Gabriela Lichtenstein and Y. Liu and S. Lowette and Jayita Lahiri and Siddharth P. Maharathy},
  title = {{Reinterpretation and preservation of data and analyses in HEP}},
  year = {2025},
  eprint = {2504.00256},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2504.00256},
  howpublished = {arXiv},
}
Figure