ATLAS $t\bar{t}$ simulation for ML-based jet flavour tagging (JetSet)

Description

Flavour-tagging — the task of identifying the flavour of jets — is essential for many physics analyses at the ATLAS experiment. This dataset, released for public use, can be used to train and evaluate machine learning models for jet flavour-tagging at ATLAS. It aims to facilitate broader interest and further development of innovative machine learning techniques to improve flavour-tagging performance.

The dataset consists of approximately 50 million events from simulated top quark pair production at a centre-of-mass energy of 13.6 TeV. It is stored in HDF5 format and contains structured event-level, jet-level, track-level and truth hadron information. This dataset is designed to be compatible with the flavour-tagging algorithm development pipeline used at ATLAS, and is supported by accompanying instructions and example configurations provided in open-source repositories.

To improve usability, the dataset is split into three mutually exclusive HDF5 files:

• mc-flavtag-ttbar-small.h5 — ~1.36 million events (~5.6 million jets)
• mc-flavtag-ttbar-medium.h5 — ~6.23 million events (~25.6 million jets)
• mc-flavtag-ttbar-large.h5 — ~41.1 million events (~168 million jets)

Downloading all three files will provide access to the complete dataset. The smaller subsets are useful for quick exploration or prototyping workflows.

Dataset characteristics

How can you use these data?

A detailed explanation of this dataset, and instructions for pre-processing, training, and evaluation workflows are provided in the accompanying GitLab repository. If this dataset is used in a publication, please cite this dataset record along with the accompanying ATLAS paper describing GN2, a ATLAS flavour-tagging algorithm with a transformer-like architecture.

Transforming Jet Flavour: Documentation and training pipeline

ATLAS GN2 paper ATLAS-FTAG-2023-05