Recasting the ATLAS search for displaced hadronic jets in the ATLAS calorimeter with additional jets or leptons using surrogate models

TL;DR

This paper validates surrogate machine-learning models that reinterpret ATLAS searches for displaced hadronic jets, enabling efficient analysis of long-lived particle events using full Run-2 data.

Contribution

It introduces and validates surrogate models for ATLAS displaced jet searches, improving reinterpretation accuracy and addressing issues in previous methods.

Findings

Surrogate models accurately reproduce original analysis results.

Validation performed using standalone and HackAnalysis frameworks.

Models respond well to various event topologies.

Abstract

This note describes the validation of a new form of re-interpretation material provided by an ATLAS search for hadronically-decaying neutral long-lived particles in association with jets or leptons, using the full Run-2 dataset. This reference ATLAS analysis provided a set of machine-learning-based "surrogate models" which return the probability of an event being selected in a given channel of the analysis, using as input truth-level kinematic information (decay position, transverse momentum and decay products of the long-lived particles). In this document, we describe the surrogate model framework in detail, and how it responds to issues identified in other re-interpretation procedures. We describe independent validations of the surrogate models' performance in reproducing the original analysis results -- first using a standalone framework and then employing the HackAnalysis framework.