Top Quark Pair Events for Heavy Flavour Tagging and Vertexing at the LHC

Simulated data from the LHC contains jets from top anti-top decays. The dataset includes 13.5 million training jets, 1.35 million validation jets, and 1.35 million test jets, each categorized as b-jets, c-jets, or light-flavoured jets. Nikita Ivvan Pond from University College London created this dataset using Pythia 8 and Delphes frameworks to model the ATLAS detector response at 13 TeV.

Use Cases

• Training classifiers to identify b-jets, c-jets, and light-flavoured jets based on jet kinematics and constituent data.
• Developing vertex reconstruction algorithms using the constituent-level 'truth_vertex_idx' variable.
• Benchmarking particle flow and jet substructure analysis techniques using the detailed constituent and truth hadron information.

Strengths

• Large scale with 13.5 million training jets and balanced classes of 4.5 million jets per flavour.
• Includes detailed truth-level information with up to 5 truth hadrons and 50 charged constituents per jet for validation.
• Kinematic distributions are resampled for the training and validation sets to ensure equivalent distributions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Data is simulated and may not perfectly reflect real detector conditions or systematic uncertainties.

Provenance

Source

University College London

Collection Method

Simulated using Pythia 8 event generator and Delphes fast detector simulation framework.