Abstract
In this paper, we present a novel method for pile-up removal of $pp$interactions using variational inference with diffusion models, called vipr.Instead of using classification methods to identify which particles are fromthe primary collision, a generative model is trained to predict theconstituents of the hard-scatter particle jets with pile-up removed. Thisresults in an estimate of the full posterior over hard-scatter jetconstituents, which has not yet been explored in the context of pile-upremoval, yielding a clear advantage over existing methods especially in thepresence of imperfect detector efficiency. We evaluate the performance of viprin a sample of jets from simulated $t\bar{t}$ events overlain with pile-upcontamination. vipr outperforms softdrop and has comparable performance topuppiml in predicting the substructure of the hard-scatter jets over a widerange of pile-up scenarios.