Speaker
Description
I present a new method for unbinned cross-section measurements and related inference problems at the LHC.
The new methodology revolves around 'refinable' machine learning of various model parameter dependencies with a particular focus on systematic effects. It shows significant performance gains in concrete applications. I will illuminate the general methodology for two realistic cases: An unbinned EFT measurement of the top quark pair production process at high pairwise mass and an unbinned cross-section measurement of the H->Tautau process. The second example also summarizes our contribution to the FAIR Universe Uncertainty Challenge.
Links:
https://www.codabench.org/competitions/2977
TTbar: https://arxiv.org/abs/2406.19076
Htautau: Submission before May 19th.