AutoSciDACT Particle Physics Dataset

This is the particle physics (jets) dataset from the paper "AutoSciDACT: Automated Scientific Discovery through Contrastive Embedding and Hypothesis Testing", presented at NeurIPS 2025 in San Diego, CA. The HDF5 files contains low-dimensional embeddings of particle jets from the JetClass dataset, obtained via supervised contrastive contrastive learning, as described in the paper. Each file corresponds to a different embedding dimensionality, as indicated in the file name. The files contain the corresponding branches: embeddings - the N-dimensional jet embeddings labels - class labels 0 - 4 (0 = QCD, 1 = ttbar, 2 = W --> qq', 3 = Z --> qq, 4 = H --> bb) jet_energy - energy jet_pt - transverse momentum jet_eta - pseudorapidity jet_phi - azimuthal angle jet_nparticles - multiplicity (number of particles in the jet) jet_sdmass - soft-drop jet mass jet_tau1 - 1-subjettiness jet_tau2 - 2-subjettiness jet_tau3 - 3-subjettiness jet_tau4 - 4-subjettiness