PhyGHT Dataset: Simulated Top Quark Pair Production for HL-LHC Pileup Mitigation

This repository contains the official dataset for the paper "PhyGHT: Physics-Guided Hypergraph Transformer for Signal Purification at the HL-LHC". It consists of simulated proton-proton collision events of top quark pair production (ttbar) designed to model the extreme conditions of the High-Luminosity LHC (HL-LHC). The dataset is specifically curated to mitigate pileup. Dataset ContentThe repository includes two distinct datasets representing different pileup (noise) scenarios:1. Standard LHC: Average interactions per bunch crossing (mu) = 60.2. Extreme HL-LHC: Average interactions per bunch crossing (mu) = 200. Each dataset contains 10,000 events, with ground-truth labels for jets and tracks. Data StructureThe data is provided in Python pickle (.pkl) format. Each file contains a nested dictionary with the following keys: jets: Feature array [N_events, N_jets, N_features] containing aggregated jet kinematics and truth labels for Energy/Mass correction. trks: Feature array [N_events, N_tracks, N_features] containing particle kinematics, impact parameters, and truth labels for track signal/noise classification. jet_trk_IDs: A mapping list that links specific tracks to their parent jets. Code AvailabilityThe official PyTorch implementation and detailed dataset walkthrough and description are available at:https://github.com/rAIson-Lab/PhyGHT