Machine-learned particle flow as a foundation model for collider physics

A machine learning model trained for particle-flow reconstruction can serve as a foundation model for collider physics, performing multiple analysis tasks while using far fewer parameters than conventional architectures, according to a paper submitted to the arXiv preprint server. The workflow from a particle collision to a physics result typically passes through a series of reconstruction steps that are modular and disconnected, with no shared representation linking low-level detector data to high-level analysis tasks [2]. Researchers have now shown that casting event reconstruction as a machine learning problem naturally produces such a shared representation [2]. The team repurposed a machine learning model trained for particle-flow reconstruction, known as MLPF, to perform three distinct analysis tasks: jet flavor identification, jet energy regression, and missing momentum regression [2]. By appending the per-particle latent representations learned during reconstruction as additional input features, the model substantially improved over baselines that use kinematic features alone [2]. A single linear layer trained using only the latent representations achieved competitive performance against state-of-the-art baseline architectures and outperformed the baseline for missing momentum regression with approximately 35 times fewer parameters [2]. These results demonstrate that the latent representations learned during reconstruction encode essential physics information needed for downstream analysis, establishing MLPF as a foundation model and offering a concrete step toward an end-to-end pipeline from detector data to physics analysis [2]. The paper was submitted to arXiv on 12 Jun 2026 [1]. arXiv is an open-access repository of electronic preprints and postprints approved for posting after moderation but not peer review, and it serves as a primary dissemination channel in fields such as high-energy physics [9]. The research appears under the High Energy Physics - Experiment category [1]. The study was shared through arXivLabs, a framework that allows collaborators to develop and share new features directly on the arXiv website, with both individuals and organizations embracing values of openness, community, excellence, and user data privacy [7]. The arXivLabs framework complements other collaborative projects and demonstrates how results are achieved when the community works together, according to arXiv Executive Director Eleonora Presani [7]. The Relativistic Heavy Ion Collider at Brookhaven National Laboratory, the only operating particle collider in the United States, ended data collection on 6 February 2026, marking the conclusion of an era for a facility that had studied quark-gluon plasma at temperatures exceeding 4 terakelvin [3]. The broader field of physics was transformed at the beginning of the 20th century by the discoveries of quantum mechanics, relativity, and atomic theory [4].