Similarity Pairing with Energy Mover's Distance for Self-Supervised Pre-Training at the LHC
Abstract
Many self-supervised methods for training foundation models at the Large Hadron Collider (LHC) rely on data augmentations to encourage the model to embed events into a representation space invariant to certain physical or detector symmetries. A common challenge arises from the large freedom in choosing a proper set of augmentations on which downstream performance depends. The implementation of augmentations involves either modifying existing events, potentially breaking the event fidelity, or simulating more event variants, which is computationally intensive. In this work, we present a data-driven method of pairing events by their similarity via the energy mover's distance (EMD), which measures how similar two events are in terms of the work required to transform one into the other. With this approach, distinct events are sampled and matched by their similarity to serve as views for learning invariance, keeping the physics content of each event intact without handcrafted distortions. We demonstrate this augmentation-free pairing method by pre-training on QCD jets via self-distillation and show that it can yield semantic jet embeddings with downstream discrimination power comparable to or better than an augmentation-based baseline.
Related
- Spatial asymmetry of neutron pairing:A distance-resolved study of 46Ti
- Charge density-supervised real-space representations for atomistic energy prediction
- An energy stable and accuracy-preserving finite volume scheme based on the SAV method with application to wall-distance computation
- Equilibrium Training of Energy-Based Models with Parallel Trajectory Tempering
- Panda Diplomacy: Foundation Model Pre-training across Particle Imaging Detectors for High Energy and Nuclear Physics
BibTeX
@misc{tsoi2026similarity,
title = {{Similarity Pairing with Energy Mover's Distance for Self-Supervised Pre-Training at the LHC}},
author = {Ho Fung Tsoi and Dylan Rankin},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.17738},
eprint = {2609.17738},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.17738},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Similarity Pairing with Energy Mover's Distance for Self-Supervised Pre-Training at the LHC
AU - Ho Fung Tsoi
AU - Dylan Rankin
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2609.17738
UR - https://arxiv.org/abs/2609.17738
AB - Many self-supervised methods for training foundation models at the Large Hadron Collider (LHC) rely on data augmentations to encourage the model to embed events into a representation space invariant to certain physical or detector symmetries. A common challenge arises from the large freedom in choosing a proper set of augmentations on which downstream performance depends. The implementation of augmentations involves either modifying existing events, potentially breaking the event fidelity, or simulating more event variants, which is computationally intensive. In this work, we present a data-driven method of pairing events by their similarity via the energy mover's distance (EMD), which measures how similar two events are in terms of the work required to transform one into the other. With this approach, distinct events are sampled and matched by their similarity to serve as views for learning invariance, keeping the physics content of each event intact without handcrafted distortions. We demonstrate this augmentation-free pairing method by pre-training on QCD jets via self-distillation and show that it can yield semantic jet embeddings with downstream discrimination power comparable to or better than an augmentation-based baseline.
ER -