1–2 Jun 2026
Bloomberg Center for Physics and Astronomy
US/Eastern timezone

Unbinned multidimensional scale factors for SBI

Not scheduled
20m
462 (Bloomberg Center for Physics and Astronomy)

462

Bloomberg Center for Physics and Astronomy

3701 San Martin Drive, Baltimore, MD 21218

Speaker

Davide Valsecchi (ETH Zurich (CH))

Description

Precision detector calibration is a fundamental prerequisite for deploying realistic Simulation-Based Inference (SBI) in high-energy physics, particularly for maximizing sensitivity in Effective Field Theory (EFT) measurements. Recent advances in probabilistic machine learning enable unbinned likelihood fits that extract continuous, multidimensional calibrations simultaneously. This talk introduces Tag & Flow, a novel extension of the classic Tag-and-Probe efficiency calibration method currently under development within the CMS collaboration. Tag-and-Flow models detector efficiencies as continuous multidimensional functions while rigorously capturing their associated uncertainties. I will outline the methodology and discuss both the prospects and challenges of integrating this approach into SBI pipelines for future physics analyses

Presentation type. Talk

Author

Davide Valsecchi (ETH Zurich (CH))

Presentation materials

There are no materials yet.