Signal dataset for Generator Based Inference (GBI) study

This is the signal dataset used in the Generator Based Inference (GBI) study. This dataset is an extended version of dataset used in LHCO Olympics 2020 (https://zenodo.org/records/4536624) by extending jet1 and jet2 masses of signal samples from one fixed pair (100 GeV, 500 GeV) to 12 times 12 pairs between 50 and 600 GeV, in a 50 GeV step size. Comparing to the previous dataset used in Incorporating Physical Priors into Weakly-Supervised Anomaly Detection (PAWS) (https://zenodo.org/records/11188685), in this dataset each mass point contains at least 100k events in the signal region to make sure the normalization factor of the likelihood ratio from the pre-trained model is constant across all signals.