Inspired by the recent application of ML-based unfolding to ATLAS data, we will unfold events from pp to Z + 2 jets production from the reco-level to the pre-detector or gen-level. We generate the events with Madgraph 5, shower and hadronization are simulated with Pythia 8.311, and detector effects are included via Delphes 3.5.0 using the default CMS card. Jets are clustered at gen-level and reco-level using an anti-kT algorithm with R=0.4 implemented in FastJet~3.3.4. We apply a set of cuts resembling the ATLAS analysis. For details on the cuts please see the associated paper. All events must pass all cuts on gen and reco level. The training set consists of 1.5M events, the test set of 400k events.