(Re)interpretation of the LHC results for new physics

29 Aug 2023, 09:00 → 1 Sept 2023, 17:00 Europe/London

PH8 (James Duff Lecture Theatre) (Durham University)

Jack Araz (IPPP - Durham University), Jonathan Butterworth (UCL), Michael Spannowsky (IPPP, Durham University), Nicholas Wardle (Imperial College London), Sabine Kraml (LPSC Grenoble)

*** A national train strike on Friday, Sep 1, risks to make travel from the workshop exceedingly difficult! We're therefore trying to adapt the workshop program, accommodating the Friday talks on Wed/Thu instead. Please regularly check the timetable for short-notice updates. ***

This is the 8th general workshop of the “Forum on the interpretation of the LHC results for BSM studies”, or LHC Reinterpretation Forum (RIF) for short. It aims to review new developments on the tools, phenomenology, and experimental sides and to prepare for the Run 3 results of the LHC. In this context, the major topics of this workshop will be: 

1. Storage and (re)usage of theoretical predictions, including event samples
2. The communication and reuse of statistical and machine-learned models
3. The combined/global interpretation of searches and measurements 

Moreover, a full day will be dedicated to EFT-related aspects in a joint session with the LHC EFT working group!

 

We will also continue the conversation on general best practices for the publication and reuse of experimental results beyond the LHC and therefore welcome contributions regarding results from, e.g., intensity-frontier or astrophysical experiments.

Background: The RIF aims to discuss topics related to the BSM (re)interpretation of LHC data, including the development of the necessary public recasting tools and related infrastructure, and to provide a platform for continuous interaction between theorists and with the experiments. So far, seven general workshops have been held at CERN, Fermilab, and Imperial College London. They resulted in the 2020 report on Reinterpretation of LHC Results for New Physics: Status and recommendations after Run 2, arXiv:2003.07868, and the white paper on Publishing statistical models: Getting the most out of particle physics experiments, arXiv:2109.04981 (both published in SciPost Physics), which will be the basis of this meeting.

(Re)interpretation LHC 2023 Credit card form

Rochester Building Ground Level Layout

Tuesday, 29 August

Welcome & Introduction

Convener: Jack Araz (IPPP - Durham University)

Reinterpretation studies / pheno

Convener: Jack Araz (IPPP - Durham University)

1 Testing the scalar triplet solution to CDF’s heavy W problem at the LHC

The type II seesaw model remains a popular and viable explanation of neutrino masses and mixing angles. By hypothesizing the existence of a scalar that is a triplet under the weak gauge interaction, the model predicts strong correlations among neutrino oscillation parameters, signals at lepton flavour experiments, and collider observables at high energies. We investigate reports that the type II seesaw can naturally accommodate recent measurements by the CDF collaboration, which finds the mass of the W boson to be significantly larger than allowed by electroweak precision data, while simultaneously evading constraints from direct searches. Experimental scrutiny of this parameter space in the type II seesaw has long been evaded since it is not characterized by “golden channels” at colliders but instead by cascade decays, moderate mass splittings, and many soft final states. In this work, we test this parameter space against publicly released measurements made at the Large Hadron Collider. By employing a newly developed tool chain combining madgraph5_amc@nlo and contur, we find that most of the favoured space for this discrepancy is already excluded by measurements of Standard Model final states. We give suggestions for further exploration at run III of the LHC, which is now under way.

Speaker: Jonathan Butterworth (UCL)

bidirectional_linking_in_hepdata.pptx

durham_rif.pptx

11:00 Caffeine break

2 Global constraints of the electroweak-ino sector of the MSSM with SModelS 2.3

Reinterpretations of LHC searches are typically done in a channel-by-channel approach, considering the constraints from each experimental search separately. In this presentation, we go a step further and discuss how combining LHC search results into a global likelihood can give better and statistically more robust constraints on the tested models. Concretely, our analysis focuses on the electroweakino sector of the MSSM, for which we derive global constraints based on the ~10 publicly available and reusable ATLAS and CMS searches for signals in this sector through the SModelS package.

Speaker: Timothee Pascal

global_constraints_on_ewino_with_smodels_RiF2023.pdf

3 Testing the Standard Model and beyond with $B\to PP$ decays

The starting point to our discussion is the "$B\to K\pi$ puzzle".
We show, that although the "puzzle" can be resolved by a more detailed analysis, there is a more fundamental question that needs to be addressed:
Is New Physics necessary to describe the experimentally observed asymmetries and branching fractions of the $B\to PP$ decays?
We perform a phenomenological analysis based on fits of an model-independent New Physics parameterisation and obtain exclusion $\chi^2$ plots of the two ad-hoc parameters (magnitude and a weak phase) sensitive to New Physics.
The results show that the Standard Model expectation value lies within $1\sigma$ with respect to the global minimum but the overall picture is not well constrained by the existing data. Our results are mostly sensitive to the time-dependent asymmetry and branching fraction of the $B^0\to K^0\pi^0$ decay. New experimental outcomes are highly expected and results of this analysis provide strong hints for improvement of precision in current experiments LHCb and Belle II.
Although the Standard Model expectations seem fully consistent with the experimental data, our approach can be used to constrain New Physics models, \emph{e.g.} leading to limits on Wilson coefficients of four-quark SMEFT operators or specific $Z'$ models.

Speaker: Adam Szabelski (National Centre for Nuclear Research)

Durham2023_Reinterpretation_workshop_AdamSzabelski.pdf

4 Determining The Sign Of \kappa_{Z} Using New Resonance Searches.

In this talk I will describe a method to determine the relative sign between the hZZ and hWW couplings using the data which are collected by the ATLAS and CMS collaborations. Using the concept of perturbative unitarity we have developed a prescription and reconstruct a phenomenological Lagrangian using FeynRule package. After that we have constructed the UFO file for this simplified model and generate relevant signals using Madgraph event simulators. We have recasted existing direct search bounds from ATLAS and CMS to improve existing constraints on 125 GeV Higgs boson couplings with SM gauge bosons.

Speaker: Agnivo Sarkar (HRI)

ReInterpretaion.pdf

12:30 Lunch

Update on reinterpretation software

Convener: Nicholas Wardle (Imperial College London)

5 GAMBIT update and results

Placeholder for results and updates from GAMBIT

Speaker: Tomas Gonzalo (KIT)

ColliderBit.pdf

6 Progress and plans for Contur

An update on recent developments and results, current activities and future plans, for the Contur (Constraints of new theories using Rivet) toolkit.

Speaker: Yoran Yeh (UCL)

2023_08_29_contur_reinterpretation_forum.pdf

7 Signal region combination in CheckMATE

I will present new features of CheckMATE, in particular implementation of multibin fits in a number of ATLAS and CMS searches. I will show an application of the method to electroweakino scenarios and discuss notable improvements in exclusion range due to CMS multijet search.

Speaker: Krzysztof Rolbiecki (University of Warsaw)

Durham-Checkmate (1).pdf

8 New developments in SModelS

Speaker: Sabine Kraml (LPSC Grenoble)

smodels_v2.3.pdf

15:40 Caffeine break

9 ADL/CutLang developments towards transparent (re)interpretation

We report recent developments in Analysis Description Language (ADL) and the runtime interpreter/framework CutLang targeting (re)interpretation studies, including integration of machine learning models and a Jupyter-based interface for plotting. We present validation efforts and results from various LHC analyses along with studies with CMS open data. We also highlight several core developments towards a more formal, robust and automated interpreter system.

Speaker: Sezen Sekmen (Kyungpook National University)

SekmenADLCLreint2308.pdf

10 Reduce, Reuse, Recycle: an end-to-end pipeline for recycling particle physics results

Searches for new physics at the Large Hadron Collider have constrained many models of physics beyond the Standard Model. Many searches also provide resources that allow them to be reinterpreted in the context of other models. We describe a reinterpretation pipeline that examines previously untested models of new physics using supplementary information from ATLAS Supersymmetry (SUSY) searches in a way that provides accurate constraints even for models that differ meaningfully from the benchmark models of the original analysis. The public analysis information, such as public analysis routines and serialized probability models, is combined with common event generation and simulation toolkits MadGraph, Pythia8, and Delphes into workflows steered by TOML configuration files, and bundled into the mapyde python package. The use of mapyde is demonstrated by constraining previously untested SUSY models with compressed sleptons and electroweakinos using ATLAS results.

Speaker: Giordon Stark (SCIPP, UC Santa Cruz)

20230829_RiF_mapyde.pdf

Wednesday, 30 August

Storage and (re)usage of theoretical predictions, including event samples

Convener: Jonathan Butterworth (UCL)

11 Pineline: Industrialization of High-Energy Theory Predictions

We present a collection of tools automating the efficient computation of large sets of theory predictions for high-energy physics. Calculating predictions for different processes often require dedicated programs. These programs, however, accept inputs and produce outputs that are usually very different from each other. The industrialization of theory predictions is achieved by a framework which harmonizes inputs (runcard, parameter settings), standardizes outputs (in the form of grids), produces reusable intermediate objects, and carefully tracks all meta data required to reproduce the computation. Parameter searches and fitting of non-perturbative objects are exemplary use cases that require a full or partial re-computation of theory predictions and will thus benefit of such a toolset.

Speaker: Andrea Barontini (University of Milan)

Barontini_Pineline.pdf

12 Deconstructing signals of new physics at collider

One of the critical points in performing phenomenological analyses of new physics at collider is that, very often, large parameter scans are necessary: intensive and often redundant MC simulations have to be performed to cover relevant regions of signal parameter space and achieve enough accuracy in the determination of signal features.
On the other hand, disk space and computing time are often limited, and the environmental impact of performing such computations is almost never taken into consideration.
There is a growing need to devise strategies to optimise data production and share resources in the HEP community, both theory and experiment.
I will describe a framework which allows such approach, where simulated signal samples are deconstructed into complete sets of basic elements to be combined a posteriori to perform different analysis.
The framework is modular, collaborative, flexible and resource-friendly.
I will describe it through a concrete example: the analysis of BSM contributions in Higgs pair production at the LHC, and indicate possible short- and long-term developments and applications.

Speaker: Luca Panizzi (Universita' della Calabria)

NPdeconstruction.pdf

11:00 Caffeine break

13 LHC data for PDF determination: challenges and prospects

I discuss how current LHC measurements are employed in the determination of the Parton Distribution Functions (PDFs) of the proton. I focus on the current praxis and I outline a few challenges in the release and usage of the experimental data. I finally discuss prospects to improve best practice measures, in light of LHC Run III and beyond.

Speaker: Emanuele Roberto Nocera (Università degli Studi di Torino and INFN Torino)

NOCERA_Durham.pdf

The NNPDF code

14 Bidirectional linking in HEPData

Updates on HEPData

Speaker: Jordan Byers (Durham University)

bidirectional_linking_in_hepdata.odp

bidirectional_linking_in_hepdata.pdf

Link to HEPData submission docs

15 Discussion

Speakers: Andy Buckley (University of Glasgow (GB)), Christian Gutschow (UCL), Jonathan Butterworth (UCL)

2023-08 Buckley RiF HD discussion.pdf

12:30 Lunch

Overcoming challenges I: Publishing and using statistical models

Convener: Sabine Kraml (LPSC Grenoble)

16 Spey: smooth inference for reinterpretation studies

Analysing statistical models is at the heart of any empirical study for hypothesis testing. We present a new cross-platform Python-based package which employs different likelihood prescriptions through a plug-in system, enabling the statistical inference of hypotheses. This framework empowers users to propose, examine, and publish new likelihood prescriptions without the need for developing a new inference system. Within this package, we propose a new simplified likelihood prescription that surpasses its predecessors' approximation accuracy by incorporating asymmetric uncertainties. Furthermore, this package facilitates the integration of various likelihood combination routines, thereby broadening the scope of independent studies through a meta-analysis. By remaining agnostic to the source of the likelihood prescription and the signal hypothesis generator, our platform allows for the seamless implementation of packages with different likelihood prescriptions, fostering compatibility and interoperability.

Speaker: Jack Araz (IPPP - Durham University)

arXiv: 2307.06996 [hep-ph]

Online documentation

Slides

Spey GitHub page

17 Using a Neural Network to Approximate the Negative Log Likelihood Function

An increasingly frequent challenge faced in HEP data analysis is to characterize the agreement between a prediction that depends on a dozen or more model parameters-such as predictions coming from an effective field theory (EFT) framework-and the observed data. Traditionally, such characterizations take the form of a negative log likelihood (NLL) distribution, which can only be evaluated numerically. The lack of a closed-form description of the NLL function makes it difficult to convey results of the statistical analysis. Typical results are limited to extracting "best fit" values of the model parameters and 1-D intervals or 2-D contours extracted from scanning the higher dimensional parameter space. It is desirable to explore these high-dimensional model parameter spaces in more sophisticated ways. One option for overcoming this challenge is to use a neural network to approximate the NLL function. This approach has the advantage of being continuous and differentiable by construction, which are essential properties for an NLL function and may also provide useful handles in exploring the NLL as a function of the model parameters. In this talk, we demonstrate the application of this technique to an analysis involving a search for new physics in the top quark sector within the framework of effective field theory. We also touch on options for distributing this likelihood function in a portable fashion.

Speaker: Nathan Jamieson (University of Notre Dame)

LHCReinterpretation_DNN.pptx

18 A model-independent likelihood function for the Belle II $B^+\to K^{+} \nu \bar{\nu}$ analysis

Rare decays like $B^+ \to K^+ \nu \bar{\nu}$, searched for by the Belle II collaboration, are important in particle physics research as they offer a window into physics beyond the Standard Model. However, the experimental challenges induced by the two final state neutrinos require assumptions on the kinematic distribution of this decay. Consequently, the results feature a model dependency arising from both Standard Model assumptions and from the description of the pertinent hadronic matrix element, making reinterpretation complicated without reanalysing the underlying data.

In this work, we address this issue by deriving a model-independent likelihood function, parameterizing the theory space in terms of Wilson coefficients of the weak effective theory, and reweighting the signal template according to the predicted kinematic signal distribution.

By extending the pyhf fitting software and interfacing it with the EOS software for flavor physics phenomenology, we can perform a runtime update of the theoretical model, enabling us to derive exclusion limits in the space of Wilson coefficients.

Once public, the model-independent likelihood function will be a useful tool for the particle physics community to perform tests on existing theoretical models.

Speaker: Lorenz Gaertner (Belle II / Ludwig-Maximilians-University Munich)

reinterpretation_forum.pdf

19 Active Learning for analysis reinterpretation and constraining additional physics parameters

Most Beyond Standard Model (BSM) physics theories are characterized by multiple BSM parameters. These encompass properties like new particle masses, coupling constants, decay widths, and effective field theory parameters.

When testing such theories against data, analysts might choose to consider only a subset of relevant BSM physics parameter in order to work within limits of computational resources and person power. Nonetheless, it is generally desirable to constrain all relevant physics parameters in a given BSM theory and produce exclusion contours in the n-dimensional space spanned by the BSM parameters.

To make the calculation of such n-dimensional exclusion contours more tenable, we present an advanced automated analysis workflow that allows the computationally efficient reinterpretation of preserved analyses.
The workflow comprises an integrated analysis pipeline and active learning. The former automatically executes all steps of an analysis from event generation through to limit setting, reducing thus the required labor.

On the other hand, active learning is employed to guide the sampling of the multi-dimensional BSM parameter phase space to find the exclusion contours in an iterative process. The sampling process focuses on prioritizing points in the vicinity of the exclusion region, while reducing the sampling density in less relevant areas. This allows exclusions over high-dimensional theory phase spaces that are otherwise impractical.

We showcase the implementation of the workflow in the Production and Distributed Analysis (PanDA) system and intelligent Data Delivery Service (iDDS) in ATLAS. To demonstrate its utility, we present its application in an extended search for a dark Z-boson using events with four-lepton final states. Additionally, we highlight the progress made towards a second demonstration in a heavy Higgs analysis.

Speaker: Christian Weber (Brookhaven National Laboratory)

2023_08_30 - ReinterpretaionWS - Active Learning - Draft 1.pptx

16:00 Caffeine break

20 Workshop photo

21 Using unbinned measurements for new physics

Machine learning tools have enabled a new type of differential cross section measurements that are unbinned and high-dimensional (see e.g. 2109.13243). This talk will discuss the challenges and prospects of (re)using such measurements with respect to new physics.

Speaker: Benjamin Nachman (Lawrence Berkeley National Laboratory)

Reinterpretation2023.pdf

Extra session

22 HEP Statistics Serialization Standard

The HEP Statistics Serialization Standard is an emerging standard to
serialize statistical models in High Energy Physics as JSON (or
JSON-like) files.

Publishing likelihoods has been on the bucket-list of the HEP
community since 20 years. With this new standard, which is already
implemented in ROOT and has gathered support by all major HEP
statistics frameworks, the community will finally and for the first
time have the option to publish full statistical models of LHC
measurements in a software-independent format. The HS3 standard will
not only allow for long-term archiving of analyses but also facilitate
the combination and reinterpretation of LHC measurements.

The project is available on GitHub and overleaf and is actively
developed by physicists and software engineers from different
collaborations and statistical toolkits. Feedback and suggestions for
improvements are highly welcome.

Speaker: Carsten Burgard (TU-Dortmund)

slides

Thursday, 31 August

Extra session

Convener: Jack Araz (IPPP - Durham University)

23 Recastable ML: guidelines, surrogate models and all that

Speaker: Tomasz Procter (University of Glasgow)

NN_REINTERPRET_DURHAM_23.pdf

24 Parametrising profiled likelihoods with neural networks

Full statistical models encapsulate the complete information of an experimental result, including the likelihood function given observed data. Since a few years ago ATLAS started publishing statistical models that can be reused via the pyhf framework; a major step towards fully publishing LHC results. In the case of fast Simplified Model Spectra based reinterpretation we are often only interested in the profiled likelihood given a signal strength. However, their computation using pyhf take the order of seconds per parameter point, slowing down SMS reinterpretation by orders of magnitude. Thus, to fully leverage from the precision obtained from full statistical models without compromising speed, we propose to learn the profiled likelihood functions with Neural Networks (NNs). We show that such functions can be well described with simple NNs, that can be easily published in the ONNX format.

Speaker: Humberto Reyes-Gonzalez

MLProfLike.pdf

EFT fits

Convener: Ken Mimasu

25 Off-the-shelf flavour constraints for your SMEFT fit

I will discuss how SMEFT fits incorporate constraints from quark flavour physics
at the hand of two recent examples. I will also highlight the technical
complexity of this endeavour, due to the large number of low-energy hadronic parameters involved.
I propose an approach that "preserves" the flavourful results of such low-energy EFT fits
and provides them to subsequent SMEFT fits as "off the shelf goods".

Speaker: Dr Danny van Dyk (IPPP Durham)

RIF2023-dvandyk.pdf

26 Interpreting HEP data in SMEFiT

Global fits to particle physics data outside experimental collaborations requires the combination and (re)interpretation of a vast range of LHC datasets. This is a non trivial exercise and requires state of the art Monte Carlo simulations at NLO QCD in the Standard Model Effective Field Theory (SMEFT). In this talk, I will present the latest ongoing efforts from the SMEFiT collaboration, focusing especially on the combined top, Higgs, diboson and EW sectors. I will also show results related to explicit UV complete models via matching. Throughout the talk, I will highlight several aspects that are relevant to the interplay between theory and experiment in the context of global EFT fits.

Speaker: Jaco ter Hoeve (Nikhef / VU Amsterdam)

RIF2023_terHoeve.pdf

11:00 Caffeine break

27 Save the EFT: a primer for the ATLAS+CMS combination in the top sector

The talk includes an overview of the latest EFT studies in the top sector from the CMS experiment. Various physics processes and experimental approaches are reviewed, with an additional discussion of prospects for a potential ATLAS-CMS EFT combination for top-related processes.

Speaker: Kirill Skovpen (Ghent University)

kskovpenSaveEFT.pdf

28 Global EFT fits within the ATLAS experiment

With detailed precision measurements of physics at the scale of electroweak symmetry breaking (EWSB), the experimental study of the Standard Model at the Large Hadron Collider is rapidly shifting towards an effective field theory (EFT) approach. An effective field theory prescription allows to consistently account for all possible deformation to the Standard Model arising from decoupled New Physics scenarios. This calls for a global approach consisting of measurements from different physics sectors and experiments to uncover deviations in data.

In this talk, I will discuss a global EFT fit within ATLAS, which combines kinematic measurements of the Higgs boson and the electroweak sector and includes constraints of electroweak precision observables from LEP and SLD. I will emphasise further how such fits can be allowed to go towards more global fits both within the collaboration and also serves as an input to global fits performed outside the collaboration.

Speaker: Rahul Balasubramanian (Nikhef)

20230831_atlas_lhceftwg.pdf

12:30 Lunch

LHC EFT WG

Convener: Nicholas Wardle (Imperial College London)

29 *CANCELLED* Simultaneous determination of SMEFT coefficients and PDFs

I present a tool for the simultaneous determination of Parton Distribution Functions and EFT coefficients. I discuss the importance of this simultaneous fit, as well as challenges in obtaining proper constraints. I present the most comprehensive analysis of both PDF and EFT effects in the top sector, which will be reproducible with our tool.

Speaker: Zahari Dimitrov Kassabov Zaharieva (University of Turin)

30 Summary of LHC EFT WG activities

I will give an overview of the LHC EFT working group and summarise ongoing activities.

Speaker: Ken Mimasu

Mimasu_RIF_flat.pdf

Mimasu_RIF.pdf

31 EFTs, models and matching: the necessity and caveats

In this talk, I will address the distinction between two different kinds of Effective Field Theories. I will specifically discuss the differences between the Standard Model Effective Field Theory (SMEFT) and the Higgs Effective Field Theory and motivate the importance of both while interpreting data, providing an example for the scalar sector. I will then focus on the importance of global fits in matching a class of single-scalar extended models to SMEFT operators, giving emphasis to various measurements. Furthermore, I will review the importance of EFT-model matching in obtaining the cut-off scale during event generation. Finally, I will discuss some bounds relating Wilson Coefficients of four-fermion operators giving rise to flavour violating and flavour conserving processes.

Speaker: Shankha Banerjee (CERN)

Banerjee_Reinterpretation.pdf

32 Global view on SMEFT interpretations and UV connection

Speaker: Maeve Madigan (University of Heidelberg)

ReinterpretationForum_MM.pdf

16:20 Caffeine break

33 Discussion session on EFT (re)interpretations

Speakers: Ken Mimasu, Nicholas Wardle (Imperial College London)

34 Final words while everyone is here

Speaker: Jack Araz (IPPP - Durham University)

Attendance Stats

Friday, 1 September

Open discussion session

35 Informal discussions and hands-on

11:00 Coffee Break

36 Informal discussions and hands-on -cont-

12:30 Lunch

Open discussion session

Convener: Jack Araz (IPPP - Durham University)

37 more discussion (for those who are still around)

Speaker: Jack Araz (IPPP - Durham University)

16:00 Coffee break