<p>Resonant anomaly detection methods have great potential for enhancing the sensitivity of traditional bump hunt searches. A key component of these methods is a high quality background template used to produce an anomaly score. Using the LHC Olympics R&D dataset, we demonstrate that this background template can also be repurposed to directly estimate the background expectation in a simple cut and count setup. In contrast to a traditional bump hunt, no fit to the invariant mass distribution is needed, thereby avoiding the potential problem of background sculpting. Furthermore, direct background estimation allows working with large background rejection rates, where resonant anomaly detection methods typically show their greatest improvement in significance.</p>

United States

Ranit Das

Germany

Thorben Finke

Marie Hein

Gregor Kasieczka

Michael Krämer

Alexander Mück

David Shih

Accurate and robust methods for direct background estimation in resonant anomaly detection

Weakly supervised methods have emerged as a powerful tool for model-agnostic anomaly detection at the Large Hadron Collider (LHC). While these methods have shown remarkable performance on specific signatures such as dijet resonances, their application in a more model-agnostic manner requires dealing with a larger number of potentially noisy input features. In this paper, we show that using boosted decision trees as classifiers in weakly supervised anomaly detection gives superior performance compared to deep neural networks. Boosted decision trees are well known for their effectiveness in tabular data analysis. Our results show that they not only offer significantly faster training and evaluation times, but they are also robust to a large number of noisy input features. By using advanced gradient boosted decision trees in combination with ensembling techniques and an extended set of features, we significantly improve the performance of weakly supervised methods for anomaly detection at the LHC. This advance is a crucial step toward a more model-agnostic search for new physics.

Parada Prangchaikul

Tobias Quadfasel

Manuel Sommerhalder

Tree-based algorithms for weakly supervised anomaly detection

Transformers have become the primary architecture for natural language processing. In this study, we explore their use for auto-regressive density estimation in high-energy jet physics, which involves working with a high-dimensional space. We draw an analogy between sentences and words in natural language and jets and their constituents in high-energy physics. Specifically, we investigate density estimation for light QCD jets and hadronically decaying boosted top jets. Since transformers allow easy sampling from learned densities, we exploit their generative capability to assess the quality of the density estimate. Our results indicate that the generated data samples closely resemble the original data, as evidenced by the excellent agreement of distributions such as particle multiplicity or jet mass. Furthermore, the generated samples are difficult to distinguish from the original data, even by a powerful supervised classifier. Given their exceptional data processing capabilities, transformers could potentially be trained directly on the massive LHC data sets to learn the probability densities in high-energy jet physics.

Jan Tönshoff

Learning the language of QCD jets with transformers

Defect characterizes the depth of factorization of terms in differential (cyclotomic) expansions of knot polynomials, i.e. of the non-perturbative Wilson averages in the Chern-Simons theory. We prove the conjecture that the defect can be alternatively described as the degree in  $$q^{\pm 2}$$  <math> <msup> <mi>q</mi> <mrow> <mo>±</mo> <mn>2</mn> </mrow> </msup> </math>   of the fundamental Alexander polynomial, which formally corresponds to the case of no colors. We also pose a question if these Alexander polynomials can be arbitrary integer polynomials of a given degree. A first attempt to answer the latter question is a preliminary analysis of antiparallel descendants of the 2-strand torus knots, which provide a nice set of examples for all values of the defect. The answer turns out to be positive in the case of defect zero knots, what can be observed already in the case of twist knots. This proved conjecture also allows us to provide a complete set of C-polynomials for the symmetrically colored Alexander polynomials for defect zero. In this case, we achieve a complete separation of representation and knot variables.

Russia

E. Lanina

A. Morozov

Defect and degree of the Alexander polynomial

We show how weakly supervised machine learning can improve the sensitivity of LHC mono-jet searches to new physics models with anomalous jet dynamics. The Classification Without Labels (CWoLa) method is used to extract all the information available from low-level detector information without any reference to specific new physics models. For the example of a strongly interacting dark matter model, we employ simulated data to show that the discovery potential of an existing generic search can be boosted considerably.

Maximilian Lipp

Boosting mono-jet searches with model-agnostic machine learning

In this paper we study the group theoretic structures of colored HOMFLY polynomials in a specific limit. The group structures arise in the perturbative expansion of <math><mi>S</mi><mi>U</mi><mo>(</mo><mi>N</mi><mo>)</mo></math> Chern-Simons Wilson loops, while the limit is <math><mi>N</mi><mo>→</mo><mn>0</mn></math>. The result of the paper is twofold. First, we explain the emergence of Kadomsev-Petviashvily (KP) τ-functions. This result is an extension of what we did in [1], where a symbolic correspondence between KP equations and group factors was established. In this paper we prove that integrability of the colored Alexander polynomial is due to it's relation to soliton τ-functions. Mainly, the colored Alexander polynomial is embedded in the action of the KP generating function on the soliton τ-function. Secondly, we use this correspondence to provide a rather simple combinatoric description of the group factors in term of Young diagrams, which is otherwise described in terms of chord diagrams, where no simple description is known. This is a first step providing an explicit description of the group theoretic data of Wilson loops, which would effectively reduce them to a purely topological quantity, mainly to a collection of Vassiliev invariants.

V. Mishnyakov

A. Sleptsov

Perturbative analysis of the colored Alexander polynomial and KP soliton -functions

Autoencoders are widely used in machine learning applications, in particular for anomaly detection. Hence, they have been introduced in high energy physics as a promising tool for model-independent new physics searches. We scrutinize the usage of autoencoders for unsupervised anomaly detection based on reconstruction loss to show their capabilities, but also their limitations. As a particle physics benchmark scenario, we study the tagging of top jet images in a background of QCD jet images. Although we reproduce the positive results from the literature, we show that the standard autoencoder setup cannot be considered as a model-independent anomaly tagger by inverting the task: due to the sparsity and the specific structure of the jet images, the autoencoder fails to tag QCD jets if it is trained on top jets even in a semi-supervised setup. Since the same autoencoder architecture can be a good tagger for a specific example of an anomaly and a bad tagger for a different example, we suggest improved performance measures for the task of model-independent anomaly detection. We also improve the capability of the autoencoder to learn non-trivial features of the jet images, such that it is able to achieve both top jet tagging and the inverse task of QCD jet tagging with the same setup. However, we want to stress that a truly model-independent and powerful autoencoder-based unsupervised jet tagger still needs to be developed.

Alessandro Morandini

Ivan Oleksiyuk

Autoencoders for unsupervised anomaly detection in high energy physics

We describe predictions for top quark pair differential distributions at hadron colliders, by combining the next-to-next-to-leading order quantum chromodynamics calculations and next-to-leading order electroweak corrections with double resummation at the next-to-next-to-leading logarithmic accuracy of threshold logarithms and small-mass logarithms. To the best of our knowledge, this is the first study to present such a combination, which incorporates all known perturbative information. Numerical results are presented for the invariant-mass distribution, transverse-momentum distribution, and rapidity distributions.

Michał Czakon

Andrea Ferroglia

United Kingdom

Alexander Mitov

Davide Pagani

Andrew S. Papanastasiou

Benjamin D. Pecjak

Netherlands

Darren J. Scott

Sweden

Ioannis Tsinikos

Xing Wang

China

Li Lin Yang

Italy

Marco Zaro

Top quark pair production at complete NLO accuracy with NNLO+NNLL′ corrections in QCD L. L. Yang and X. Wang are supported in part by the National Natural Science Foundation of China (11975030, 11635001, 11575004). X. Wang is also supported in part by the Cluster of Excellence PRISMA+ (project ID 39083149). D. J. S. is supported under the ERC grant ERC-STG2015-677323. The work of D. P. and I. T. is supported by the Alexander von Humboldt Foundation, in the framework of the Sofja Kovalevskaja Award Project "Event Simulation for the Large Hadron Collider at High Precision". D. P. is also supported by the Deutsche Forschungsgemeinschaft (DFG) under Germany’s Excellence Strategy – EXC 2121"Quantum Univers" – 390833306. I. T. is also supported by the Swedish Research Council under contract number 2016-05996. The research of A. M. and A. P. has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (683211) as well as from UK STFC grants ST/L002760/1 and ST/K004883/1. The work of M. C. was supported in part by a grant of the BMBF and by the Deutsche Forschungs-gemeinschaft under grant 396021762 - TRR 257
						

We constrain the Higgs-portal model employing the vector-boson fusion channel at the LHC. In particular, we include the phenomenologically interesting parameter region near the Higgs resonance, where the Higgs-boson mass is close to the threshold for dark-matter production and a running-width prescription has to be employed for the Higgs- boson propagator. Limits for the Higgs-portal coupling as a function of the dark-matter mass are derived from the CMS search for invisible Higgs-boson decays in vector-boson fusion at 13 TeV. Furthermore, we perform projections for the 14 TeV HL-LHC and the 27 TeV HE-LHC taking into account a realistic estimate of the systematic uncertainties. The respective upper limits on the invisible branching ratio of the Higgs boson reach a level of 2% and constrain perturbative Higgs-portal couplings up to dark-matter masses of about 110 GeV.

Belgium

Jan Heisig

Eric Madge

Probing Higgs-portal dark matter with vector-boson fusion

New Z′ gauge bosons arise in many extensions of the Standard Model and predict resonances in the dilepton invariant mass spectrum. Searches for such resonances therefore provide important constraints on many models of new physics, but the resulting bounds are often calculated without interference effects. In this work we show that the effect of interference is significant and cannot be neglected whenever the Z′ width is large (for example because of an invisible contribution). To illustrate this point, we implement and validate the most recent 139 fb −1 dilepton search from ATLAS and obtain exclusion limits on general Z′ models as well as on simplified dark matter models with spin-1 mediators. We find that interference can substantially strengthen the bound on the Z′ couplings and push exclusion limits for dark matter simplified models to higher values of the Z′ mass. Together with this study we release the open-source code ZPEED, which provides fast likelihoods and exclusion bounds for general Z′ models.

Felix Kahlhoefer

Stefan Schulte

Patrick Tunney

Interference effects in dilepton resonance searches for Z′ bosons and dark matter mediators