Vinicius Massami Mikuni

A new paradigm for data-driven, model-agnostic new physics searches at colliders is emerging, and aims to leverage recent breakthroughs in anomaly detection and machine learning. In order to develop and benchmark new anomaly detection methods within this framework, it is essential to have standard datasets. To this end, we have created the LHC Olympics 2020, a community challenge accompanied by a set of simulated collider events. Participants in these Olympics have developed their methods using an R&D dataset and then tested them on black boxes: datasets with an unknown anomaly (or not). Methods made use of modern machine learning tools and were based on unsupervised learning (autoencoders, generative adversarial networks, normalizing flows), weakly supervised learning, and semi-supervised learning. This paper will review the LHC Olympics 2020 challenge, including an overview of the competition, a description of methods deployed in the competition, lessons learned from the experience, and implications for data analyses with future datasets as well as future colliders.

Abstract In high energy physics, graph-based implementations have the advantage of treating the input data sets in a similar way as they are collected by collider experiments. To expand on this concept, we propose a graph neural network enhanced by attention mechanisms called ABCNet. To exemplify the advantages and flexibility of treating collider data as a point cloud, two physically motivated problems are investigated: quark–gluon discrimination and pileup reduction. The former is an event-by-event classification, while the latter requires each reconstructed particle to receive a classification score. For both tasks, ABCNet shows an improved performance compared to other algorithms available.

Score-based generative models are a new class of generative algorithms that have been shown to produce realistic images even in high dimensional spaces, currently surpassing other state-of-the-art models for different benchmark categories and applications. In this work we introduce caloscore, a score-based generative model for collider physics applied to calorimeter shower generation. Three different diffusion models are investigated using the Fast Calorimeter Simulation Challenge 2022 dataset. caloscore is the first application of a score-based generative model in collider physics and is able to produce high-fidelity calorimeter images for all datasets, providing an alternative paradigm for calorimeter shower simulation.

Many particle physics datasets like those generated at colliders are described by continuous coordinates (in contrast to grid points like in an image), respect a number of symmetries (like permutation invariance), and have a stochastic dimensionality. For this reason, standard deep generative models that produce images or at least a fixed set of features are limiting. We introduce a new neural network simulation based on a diffusion model that addresses these limitations named fast point cloud diffusion. We show that our approach can reproduce the complex properties of hadronic jets from proton-proton collisions with competitive precision to other recently proposed models. Additionally, we use a procedure called progressive distillation to accelerate the generation time of our method, which is typically a significant challenge for diffusion models despite their state-of-the-art precision.

Methods for anomaly detection of new physics processes are often limited to low-dimensional spaces due to the difficulty of learning high-dimensional probability densities. Particularly at the constituent level, incorporating desirable properties such as permutation invariance and variable-length inputs becomes difficult within popular density estimation methods. In this work, we introduce a permutation-invariant density estimator for particle physics data based on diffusion models, specifically designed to handle variable-length inputs. We demonstrate the efficacy of our methodology by utilizing the learned density as a permutation-invariant anomaly detection score, effectively identifying jets with low likelihood under the background-only hypothesis. To validate our density estimation method, we investigate the ratio of learned densities and compare to those obtained by a supervised classification algorithm.

Physics beyond the Standard Model that is resonant in one or more dimensions has been a longstanding focus of countless searches at colliders and beyond. Recently, many new strategies for resonant anomaly detection have been developed, where sideband information can be used in conjunction with modern machine learning, in order to generate synthetic datasets representing the Standard Model background. Until now, this approach was only able to accommodate a relatively small number of dimensions, limiting the breadth of the search sensitivity. Using recent innovations in point cloud generative models, we show that this strategy can also be applied to the full phase space, using all relevant particles for the anomaly detection. As a proof of principle, we show that the signal from the R dataset from the LHC Olympics is findable with this method, opening up the door to future studies that explore the interplay between depth and breadth in the representation of the data for anomaly detection.

Abstract Diffusion generative models are promising alternatives for fast surrogate models, producing high-fidelity physics simulations. However, the generation time often requires an expensive denoising process with hundreds of function evaluations, restricting the current applicability of these models in a realistic setting. In this work, we report updates on the CaloScore architecture, detailing the changes in the diffusion process, which produces higher quality samples, and the use of progressive distillation, resulting in a diffusion model capable of generating new samples with a single function evaluation. We demonstrate these improvements using the Calorimeter Simulation Challenge 2022 dataset.

We propose a new method for Unsupervised clustering in particle physics named UCluster, where information in the embedding space created by a neural network is used to categorise collision events into different clusters that share similar properties. We show how this method can be applied to an unsupervised multiclass classification as well as for anomaly detection, which can be used for new physics searches.

Abstract Machine learning tools have empowered a qualitatively new way to perform differential cross section measurements whereby the data are unbinned, possibly in many dimensions. Unbinned measurements can enable, improve, or at least simplify comparisons between experiments and with theoretical predictions. Furthermore, many-dimensional measurements can be used to define observables after the measurement instead of before. There is currently no community standard for publishing unbinned data. While there are also essentially no measurements of this type public, unbinned measurements are expected in the near future given recent methodological advances. The purpose of this paper is to propose a scheme for presenting and using unbinned results, which can hopefully form the basis for a community standard to allow for integration into analysis workflows. This is foreseen to be the start of an evolving community dialogue, in order to accommodate future developments in this field that is rapidly evolving.

Methods for processing point cloud information have seen a great success in collider physics applications. One recent breakthrough in machine learning is the usage of Transformer networks to learn semantic relationships between sequences in language processing. In this work, we apply a modified Transformer network called Point Cloud Transformer as a method to incorporate the advantages of the Transformer architecture to an unordered set of particles resulting from collision events. To compare the performance with other strategies, we study jet-tagging applications for highly-boosted particles.

There is a growing need for anomaly detection methods that can broaden the search for new particles in a model-agnostic manner. Most proposals for new methods focus exclusively on signal sensitivity. However, it is not enough to select anomalous events - there must also be a strategy to provide context to the selected events. We propose the first complete strategy for unsupervised detection of non-resonant anomalies that includes both signal sensitivity and a data-driven method for background estimation. Our technique is built out of two simultaneously-trained autoencoders that are forced to be decorrelated from each other. This method can be deployed offline for non-resonant anomaly detection and is also the first complete online-compatible anomaly detection strategy. We show that our method achieves excellent performance on a variety of signals prepared for the ADC2021 data challenge.

There is a growing need for machine-learning-based anomaly detection strategies to broaden the search for beyond-the-Standard-Model physics at the Large Hadron Collider (LHC) and elsewhere. The first step of any anomaly detection approach is to specify observables and then use them to decide on a set of anomalous events. One common choice is to select events that have low probability density. It is a well-known fact that probability densities are not invariant under coordinate transformations, so the sensitivity can depend on the initial choice of coordinates. The broader machine learning community has recently connected coordinate sensitivity with anomaly detection and our goal is to bring awareness of this issue to the growing high-energy physics literature on anomaly detection. In addition to analytical explanations, we provide numerical examples from simple random variables and from the LHC Olympics dataset that show how using probability density as an anomaly score can lead to events being classified as anomalous or not depending on the coordinate frame.

Many particle physics datasets like those generated at colliders are described by continuous coordinates (in contrast to grid points like in an image), respect a number of symmetries (like permutation invariance), and have a stochastic dimensionality. For this reason, standard deep generative models that produce images or at least a fixed set of features are limiting. We introduce a new neural network simulation based on a diffusion model that addresses these limitations named Fast Point Cloud Diffusion (FPCD). We show that our approach can reproduce the complex properties of hadronic jets from proton-proton collisions with competitive precision to other recently proposed models. Additionally, we use a procedure called progressive distillation to accelerate the generation time of our method, which is typically a significant challenge for diffusion models despite their state-of-the-art precision.

A new paradigm for data-driven, model-agnostic new physics searches at colliders is emerging, and aims to leverage recent breakthroughs in anomaly detection and machine learning. In order to develop and benchmark new anomaly detection methods within this framework, it is essential to have standard datasets. To this end, we have created the LHC Olympics 2020, a community challenge accompanied by a set of simulated collider events. Participants in these Olympics have developed their methods using an R&D dataset and then tested them on black boxes: datasets with an unknown anomaly (or not). This paper will review the LHC Olympics 2020 challenge, including an overview of the competition, a description of methods deployed in the competition, lessons learned from the experience, and implications for data analyses with future datasets as well as future colliders.

Score based generative models are a new class of generative models that have been shown to accurately generate high dimensional calorimeter datasets. Recent advances in generative models have used images with 3D voxels to represent and model complex calorimeter showers. Point clouds, however, are likely a more natural representation of calorimeter showers, particularly in calorimeters with high granularity. Point clouds preserve all of the information of the original simulation, more naturally deal with sparse datasets, and can be implemented with more compact models and data files. In this work, two state-of-the-art score based models are trained on the same set of calorimeter simulation and directly compared.

Machine learning-based unfolding has enabled unbinned and high-dimensional differential cross section measurements. Two main approaches have emerged in this research area; one based on discriminative models and one based on generative models. The main advantage of discriminative models is that they learn a small correction to a starting simulation while generative models scale better to regions of phase space with little data. We propose to use Schr\"odinger bridges and diffusion models to create sbunfold, an unfolding approach that combines the strengths of both discriminative and generative models. The key feature of sbunfold is that its generative model maps one set of events into another without having to go through a known probability density as is the case for normalizing flows and standard diffusion models. We show that sbunfold achieves excellent performance compared to state of the art methods on a synthetic $Z+\text{jets}$ dataset.

Abstract Knowledge distillation is a form of model compression that allows artificial neural networks of different sizes to learn from one another. Its main application is the compactification of large deep neural networks to free up computational resources, in particular on edge devices. In this article, we consider proton-proton collisions at the High-Luminosity LHC (HL-LHC) and demonstrate a successful knowledge transfer from an event-level graph neural network (GNN) to a particle-level small deep neural network (DNN). Our algorithm, \DN, is a DNN that is trained to learn about the provenance of particles, as provided by the soft labels that are the GNN outputs, to predict whether or not a particle originates from the primary interaction vertex. The results indicate that for this problem, which is one of the main challenges at the HL-LHC, there is minimal loss during the transfer of knowledge to the small student network, while improving significantly the computational resource needs compared to the teacher. This is demonstrated for the distilled student network on a CPU, as well as for a quantized and pruned student network deployed on an FPGA. Our study proves that knowledge transfer between networks of different complexity can be used for fast artificial intelligence (AI) in high-energy physics that improves the expressiveness of observables over non-AI-based reconstruction algorithms. Such an approach can become essential at the HL-LHC experiments, e.g., to comply with the resource budget of their trigger stages.

Machine learning has become an essential tool in jet physics. Due to their complex, high-dimensional nature, jets can be explored holistically by neural networks in ways that are not possible manually. However, innovations in all areas of jet physics are proceeding in parallel. We show that specially constructed machine learning models trained for a specific jet classification task can improve the accuracy, precision, or speed of all other jet physics tasks. This is demonstrated by training on a particular multiclass classification task and then using the learned representation for different classification tasks, for datasets with a different (full) detector simulation, for jets from a different collision system ($pp$ versus $ep$), for generative models, for likelihood ratio estimation, and for anomaly detection. Our OmniLearn approach is thus a foundation model and is made publicly available for use in any area where state-of-the-art precision is required for analyses involving jets and their substructure.

Recent innovations from machine learning allow for data unfolding, without binning and including correlations across many dimensions. We describe a set of known, upgraded, and new methods for ML-based unfolding. The performance of these approaches are evaluated on the same two datasets. We find that all techniques are capable of accurately reproducing the particle-level spectra across complex observables. Given that these approaches are conceptually diverse, they offer an exciting toolkit for a new class of measurements that can probe the Standard Model with an unprecedented level of detail and may enable sensitivity to new phenomena.

The Electron-Ion Collider (EIC), a state-of-the-art facility for studying the strong force, is expected to begin commissioning its first experiments in 2028. This is an opportune time for artificial intelligence (AI) to be included from the start at this facility and in all phases that lead up to the experiments. The second annual workshop organized by the AI4EIC working group, which recently took place, centered on exploring all current and prospective application areas of AI for the EIC. This workshop is not only beneficial for the EIC, but also provides valuable insights for the newly established ePIC collaboration at EIC. This paper summarizes the different activities and R&D projects covered across the sessions of the workshop and provides an overview of the goals, approaches and strategies regarding AI/ML in the EIC community, as well as cutting-edge techniques currently studied in other experiments.

Abstract Score based generative models are a new class of generative models that have been shown to accurately generate high dimensional calorimeter datasets. Recent advances in generative models have used images with 3D voxels to represent and model complex calorimeter showers. Point clouds, however, are likely a more natural representation of calorimeter showers, particularly in calorimeters with high granularity. Point clouds preserve all of the information of the original simulation, more naturally deal with sparse datasets, and can be implemented with more compact models and data files. In this work, two state-of-the-art score based models are trained on the same set of calorimeter simulation and directly compared.

A measurement of the Higgs boson Yukawa coupling to the top quark is presented using proton-proton collision data at $\sqrt{s} =$ 13 TeV, corresponding to an integrated luminosity of 137 fb$^{-1}$, recorded with the CMS detector. The coupling strength with respect to the standard model value, $Y_\mathrm{t}$, is determined from kinematic distributions in $\mathrm{t\bar{t}}$ final states containing ee, $μμ$, or e$μ$ pairs. Variations of the Yukawa coupling strength lead to modified distributions for $\mathrm{t\bar{t}}$ production. In particular, the distributions of the mass of the $\mathrm{t\bar{t}}$ system and the rapidity difference of the top quark and antiquark are sensitive to the value of $Y_\mathrm{t}$. The measurement yields a best fit value of $Y_\mathrm{t} =$ 1.16 $^{+0.24}_{-0.35}$, bounding $Y_\mathrm{t}$ $\lt$ 1.54 at a 95% confidence level.

The computational cost for high energy physics detector simulation in future experimental facilities is going to exceed the current available resources. To overcome this challenge, new ideas on surrogate models using machine learning methods are being explored to replace computationally expensive components. Additionally, differentiable programming has been proposed as a complementary approach, providing controllable and scalable simulation routines. In this document, new and ongoing efforts for surrogate models and differential programming applied to detector simulation are discussed in the context of the 2021 Particle Physics Community Planning Exercise (`Snowmass').

Diffusion generative models are promising alternatives for fast surrogate models, producing high-fidelity physics simulations. However, the generation time often requires an expensive denoising process with hundreds of function evaluations, restricting the current applicability of these models in a realistic setting. In this work, we report updates on the CaloScore architecture, detailing the changes in the diffusion process, which produces higher quality samples, and the use of progressive distillation, resulting in a diffusion model capable of generating new samples with a single function evaluation. We demonstrate these improvements using the Calorimeter Simulation Challenge 2022 dataset.

A search for exclusive two-photon production via photon exchange in proton-proton collisions, pp $\to$ p$γγ$p with intact protons, is presented. The data correspond to an integrated luminosity of 9.4 fb$^{-1}$ collected in 2016 using the CMS and TOTEM detectors at a center-of-mass energy of 13 TeV at the LHC. Events are selected with a diphoton invariant mass above 350 GeV and with both protons intact in the final state, to reduce backgrounds from strong interactions. The events of interest are those where the invariant mass and rapidity calculated from the momentum losses of the forward-moving protons matches the mass and rapidity of the central, two-photon system. No events are found that satisfy this condition. Interpreting this result in an effective dimension-8 extension of the standard model, the first limits are set on the two anomalous four-photon coupling parameters. If the other parameter is constrained to its standard model value, the limits at 95% CL are $\lvertζ_1\rvert$ $\lt$ 2.9 $\times$ 10$^{-13}$ GeV$^{-4}$ and $\lvertζ_2\rvert$ $\lt$ 6.0 $\times$ 10$^{-13}$ GeV$^{-4}$.

Methods for anomaly detection of new physics processes are often limited to low-dimensional spaces due to the difficulty of learning high-dimensional probability densities. Particularly at the constituent level, incorporating desirable properties such as permutation invariance and variable-length inputs becomes difficult within popular density estimation methods. In this work, we introduce a permutation-invariant density estimator for particle physics data based on diffusion models, specifically designed to handle variable-length inputs. We demonstrate the efficacy of our methodology by utilizing the learned density as a permutation-invariant anomaly detection score, effectively identifying jets with low likelihood under the background-only hypothesis. To validate our density estimation method, we investigate the ratio of learned densities and compare to those obtained by a supervised classification algorithm.

Machine learning-based unfolding has enabled unbinned and high-dimensional differential cross section measurements. Two main approaches have emerged in this research area: one based on discriminative models and one based on generative models. The main advantage of discriminative models is that they learn a small correction to a starting simulation while generative models scale better to regions of phase space with little data. We propose to use Schroedinger Bridges and diffusion models to create SBUnfold, an unfolding approach that combines the strengths of both discriminative and generative models. The key feature of SBUnfold is that its generative model maps one set of events into another without having to go through a known probability density as is the case for normalizing flows and standard diffusion models. We show that SBUnfold achieves excellent performance compared to state of the art methods on a synthetic Z+jets dataset.

Machine learning-based simulations, especially calorimeter simulations, are promising tools for approximating the precision of classical high energy physics simulations with a fraction of the generation time. Nearly all methods proposed so far learn neural networks that map a random variable with a known probability density, like a Gaussian, to realistic-looking events. In many cases, physics events are not close to Gaussian and so these neural networks have to learn a highly complex function. We study an alternative approach: Schr\"{o}dinger bridge Quality Improvement via Refinement of Existing Lightweight Simulations (SQuIRELS). SQuIRELS leverages the power of diffusion-based neural networks and Schr\"{o}dinger bridges to map between samples where the probability density is not known explicitly. We apply SQuIRELS to the task of refining a classical fast simulation to approximate a full classical simulation. On simulated calorimeter events, we find that SQuIRELS is able to reproduce highly non-trivial features of the full simulation with a fraction of the generation time.

Physics beyond the Standard Model that is resonant in one or more dimensions has been a longstanding focus of countless searches at colliders and beyond. Recently, many new strategies for resonant anomaly detection have been developed, where sideband information can be used in conjunction with modern machine learning, in order to generate synthetic datasets representing the Standard Model background. Until now, this approach was only able to accommodate a relatively small number of dimensions, limiting the breadth of the search sensitivity. Using recent innovations in point cloud generative models, we show that this strategy can also be applied to the full phase space, using all relevant particles for the anomaly detection. As a proof of principle, we show that the signal from the R\&D dataset from the LHC Olympics is findable with this method, opening up the door to future studies that explore the interplay between depth and breadth in the representation of the data for anomaly detection.

We propose a novel strategy for disentangling proton collisions at hadron colliders such as the LHC that considerably improves over the current state of the art. Employing a metric inspired by optimal transport problems as the cost function of a graph neural network, our algorithm is able to compare two particle collections with different noise levels and learns to flag particles originating from the main interaction amidst products from up to 200 simultaneous pileup collisions. We thereby sidestep the critical task of obtaining a ground truth by labeling particles and avoid arduous human annotation in favor of labels derived in situ through a self-supervised process. We demonstrate how our approach---which, unlike competing algorithms, is trivial to implement---improves the resolution in key objects used in precision measurements and searches alike and present large sensitivity gains in searching for exotic Higgs boson decays at the High-Luminosity LHC.

Knowledge distillation is a form of model compression that allows to transfer knowledge between intelligent algorithms. Its main application is the compactification of large deep neural networks to free up computational resources, in particular on edge devices. In this article, we consider proton-proton collisions at the High-Luminosity LHC (HL-LHC) and demonstrate a successful knowledge transfer from an event-level graph neural network (GNN) to a particle-level small deep neural network (DNN). Our algorithm, DistillNet, is a DNN that is trained to learn about the provenance of particles, as provided by the soft labels that are the GNN outputs, to predict whether or not a particle originates from the primary interaction vertex. The results indicate that for this problem, which is one of the main challenges at the HL-LHC, there is minimal loss during the transfer of knowledge to the small student network, while improving significantly the computational resource needs compared to the teacher. This is demonstrated for the distilled student network on a CPU, as well as for a quantized and pruned student network deployed on an FPGA. Our study proves that knowledge transfer between networks of different complexity can be used for fast artificial intelligence (AI) in high-energy physics that improves the expressiveness of observables over non-AI-based reconstruction algorithms. Such an approach can become essential at the HL-LHC experiments, e.g., to comply with the resource budget of their trigger stages.

Electron showers in a 10x10 calorimeter simulated with fast simulation methods (GFlash) and full simulation (Geant4). These showers were used for the Schroedinger Bridge Refinement project.<br> 4 different electron energy settings are available:<br> - randomly sampled between 10 and 100 GeV (split into a training and eval set)<br> - 20 GeV<br> - 50 GeV<br> - 80 GeV Each set contains 100k showers. Each shower is a 201-dimensional object, where the first 100 dimensions correspond to the deposited energies in the 10x10 calorimeter cells, the next 100 dimensions. describe the total number of depositions in the cells (not used in the refinement project) and the final dimension gives the incident electron energy. The energy depositions are in units of [GeV], the incident particle energy in units of [MeV]

Electron showers in a 10x10 calorimeter simulated with fast simulation methods (GFlash) and full simulation (Geant4). These showers were used for the Schroedinger Bridge Refinement project.<br> 4 different electron energy settings are available:<br> - randomly sampled between 10 and 100 GeV (split into a training and eval set)<br> - 20 GeV<br> - 50 GeV<br> - 80 GeV Each set contains 100k showers. Each shower is a 201-dimensional object, where the first 100 dimensions correspond to the deposited energies in the 10x10 calorimeter cells, the next 100 dimensions. describe the total number of depositions in the cells (not used in the refinement project) and the final dimension gives the incident electron energy. The energy depositions are in units of [GeV], the incident particle energy in units of [MeV]

A search for long-lived particles (LLPs) produced in association with a Z boson is presented. The study is performed using data from proton-proton collisions with a center-of-mass energy of 13 TeV recorded by the CMS experiment during 2016-2018, corresponding to an integrated luminosity of 117 fb$^{-1}$. The LLPs are assumed to decay to a pair of standard model quarks that are identified as displaced jets within the CMS tracker system. Triggers and selections based on Z boson decays to electron or muon pairs improve the sensitivity to light LLPs (down to 15 GeV). This search provides sensitivity to beyond the standard model scenarios which predict LLPs produced in association with a Z boson. In particular, the results are interpreted in the context of exotic decays of the Higgs boson to a pair of scalar LLPs (H $\to$ SS). The Higgs boson decay branching fraction is constrained to values less than 6% for proper decay lengths of 10-100 mm and for LLP masses between 40 and 55 GeV. In the case of low-mass ($\approx$15 GeV) scalar particles that subsequently decay to a pair of b quarks, the search is sensitive to branching fractions $\mathcal{B}$(H $\to$ SS) $\lt$ 20% for proper decay lengths of 10-50 mm. The use of associated production with a Z boson increases the sensitivity to low-mass LLPs of this analysis with respect to gluon fusion searches. In the case of 15 GeV scalar LLPs, the improvement corresponds to a factor of 2 at a proper decay length of 30 mm.

A measurement of inclusive four-jet production in proton-proton collisions at a center-of-mass energy of 13\TeV is presented. The transverse momenta of jets within $\lvert\eta\rvert \lt$ 4.7 reach down to 35, 30, 25, and 20 GeV for the first-, second-, third-, and fourth-leading jet, respectively. Differential cross sections are measured as functions of the jet transverse momentum, jet pseudorapidity, and several other observables that describe the angular correlations between the jets. The measured distributions show sensitivity to different aspects of the underlying event, parton shower, and matrix element calculations. In particular, the interplay between angular correlations caused by parton shower and double-parton scattering contributions is shown to be important. The double-parton scattering contribution is extracted by means of a template fit to the data, using distributions for single-parton scattering obtained from Monte Carlo event generators and a double-parton scattering distribution constructed from inclusive single-jet events in data. The effective double-parton scattering cross section is calculated and discussed in view of previous measurements and of its dependence on the models used to describe the single-parton scattering background.

Score-based generative models are a new class of generative algorithms that have been shown to produce realistic images even in high dimensional spaces, currently surpassing other state-of-the-art models for different benchmark categories and applications. In this work we introduce CaloScore, a score-based generative model for collider physics applied to calorimeter shower generation. Three different diffusion models are investigated using the Fast Calorimeter Simulation Challenge 2022 dataset. CaloScore is the first application of a score-based generative model in collider physics and is able to produce high-fidelity calorimeter images for all datasets, providing an alternative paradigm for calorimeter shower simulation.

There is a growing need for machine learning-based anomaly detection strategies to broaden the search for Beyond-the-Standard-Model (BSM) physics at the Large Hadron Collider (LHC) and elsewhere. The first step of any anomaly detection approach is to specify observables and then use them to decide on a set of anomalous events. One common choice is to select events that have low probability density. It is a well-known fact that probability densities are not invariant under coordinate transformations, so the sensitivity can depend on the initial choice of coordinates. The broader machine learning community has recently connected coordinate sensitivity with anomaly detection and our goal is to bring awareness of this issue to the growing high energy physics literature on anomaly detection. In addition to analytical explanations, we provide numerical examples from simple random variables and from the LHC Olympics Dataset that show how using probability density as an anomaly score can lead to events being classified as anomalous or not depending on the coordinate frame.

We propose a novel strategy for disentangling proton collisions at hadron colliders such as the LHC that considerably improves over the current state of the art. Employing a metric inspired by optimal transport problems as the cost function of a graph neural network, our algorithm is able to compare two particle collections with different noise levels and learns to flag particles originating from the main interaction amidst products from up to 200 simultaneous pileup collisions. We thereby sidestep the critical task of obtaining a ground truth by labeling particles and avoid arduous human annotation in favor of labels derived in situ through a self-supervised process. We demonstrate how our approach - which, unlike competing algorithms, is trivial to implement - improves the resolution in key objects used in precision measurements and searches alike and present large sensitivity gains in searching for exotic Higgs boson decays at the High-Luminosity LHC.

The Alpha Magnetic Spectrometer (AMS-02) is a high-energy particle physics detector operating on the International Space Station (ISS) since May 2011.Since its launch, the AMS-02 provided a large amount of data whose precision was never before achieved, opening a new path for the study of cosmic rays (CRs).The first published results of AMS-02 1-3 show tension with the current understanding of the cosmic ray theory, particularly at higher energies.These tensions are directly linked to many fundamental questions like the dark matter nature, the CR origin and their propagation through the galaxy.This work presents the measurement of the electron flux and the positron flux in primary cosmic rays, based on the data collected between May 2011 and November 2016, an extended data set with respect to the published AMS-02 results. 3The results extend the energy range explored up to 1 TeV for electrons and up to 700 GeV for positrons, being consistent with the published results when using the same data set.A discrepancy between the new measurement and the published flux is observed in the low energy region of the electron flux, while the positron flux is in good agreement.This can be explained by a charge dependent solar modulation effect.This hypothesis was investigated by studying the time evolution of the fluxes, focusing on the energy region below 40 GeV, where an electron and positron flux is computed over 74 time bins of 27 days width, corresponding to the sun's rotation period as seen from the Earth.The time dependent analysis confirms hints of charge dependent solar modulation, that are also observed by other independent analysis that have been carried out in parallel within the collaboration.

Recent measurements performed by the CMS Collaboration regarding the inclusive and differential top quark pair production are presented.The measurements are performed using the data collected during CERN LHC Run 2, at a centre-of-mass energy of 13 TeV.Measurements are compared to state-of-the-art simulations of standard model top quark production at next-to-leading order and next-to-next-leading order for different kinematic distributions related to the top quark pair system or that uses the system as a proxy for further studies.

The strong Coulomb field created in ultrarelativistic heavy ion collisions is expected to produce a rapidity-dependent difference ($\Delta v_2$) in the second Fourier coefficient of the azimuthal distribution (elliptic flow, $v_2$) between $\mathrm{D}^0$ ($\mathrm{\bar{u}c}$) and $\overline{\mathrm{D}}^0$ ($\mathrm{u\bar{c}}$) mesons. Motivated by the search for evidence of this field, the CMS detector at the LHC is used to perform the first measurement of $\Delta v_2$. The rapidity-averaged value is found to be $\langle\Delta v_2 \rangle =$ 0.001 $\pm$ 0.001 (stat) $\pm$ 0.003 (syst) in PbPb collisions at $\sqrt{s_\mathrm{NN}} =$ 5.02 TeV. In addition, the influence of the collision geometry is explored by measuring the $\mathrm{D}^0$ and $\overline{\mathrm{D}}^0$ mesons $v_2$ and triangular flow coefficient ($v_3$) as functions of rapidity, transverse momentum ($p_\mathrm{T}$), and event centrality (a measure of the overlap of the two Pb nuclei). A clear centrality dependence of prompt $\mathrm{D}^0$ meson $v_2$ values is observed, while the $v_3$ is largely independent of centrality. These trends are consistent with expectations of flow driven by the initial-state geometry.

The CMS and ATLAS detectors will face challenging conditions after the upgrade of the LHC to the High Luminosity LHC.In particular, the granularity of the pixel detectors should increase to mitigate the effect of pileup.Two possible sensor geometries are being investigated, 50 × 50 µm 2 and 25 × 100 µm 2 , to handle these conditions.One of the main factors in choosing the pixel geometry is inter-channel charge induction or crosstalk, defined as the ratio of charge induced into neighboring pixels relative to the total charge.This charge induction will affect the data rates, position resolution, and track reconstruction efficiencies.Therefore, it should be investigated carefully.The effect of crosstalk is expected to depend on the chosen pixel geometry, threshold of the signal, and readout front-end.The readout chip in this study is RD53A, developed by the RD53 Collaboration, which is a prototype investigated by both the CMS and ATLAS collaborations implementing three different analog front-end designs.Crosstalk effects are larger for the 25 × 100 µm 2 geometry, given the larger sensor capacitance.Both have been studied in the lab through direct charge injection, and also at DESY test beam facility by charge deposition of 5.6 GeV electrons in 150 µm thick silicon pixels.The effects on the cross-talk due to varying the front-end, threshold, and the impinging position of the electrons will be presented.

Recent results from the CMS Collaboration using top quarks are presented. These results are based on partial datasets collected by the CMS Collaboration during the LHC Run 2, at a center-of-mass energy of 13 TeV. This document includes the first measurement of tt‾ production in association with charm quarks, the first direct measurement of the third generation of the CKM matrix elements, the investigation of the running of the top quark mass, search for CP violation in top quark production, measurement of the forward-backward asymmetry in tt‾ production at the LHC, and the first global approach in constraining EFT operator coefficients using top quarks.

A measurement of the top quark mass is performed using a data sample enriched with single top quark events produced in the $t$ channel. The study is based on proton-proton collision data, corresponding to an integrated luminosity of 35.9 fb$^{-1}$, recorded at $\sqrt{s}$ = 13 TeV by the CMS experiment at the LHC in 2016.Candidate events are selected by requiring an isolated high-momentum lepton (muon or electron) and exactly two jets, of which one is identified as originating from a bottom quark. Multivariate discriminants are designed to separate the signal from the background. Optimized thresholds are placed on the discriminant outputs to obtain an event sample with high signal purity. The top quark mass is found to be 172.13$^{+0.76}_{-0.77}$ GeV, where the uncertainty includes both the statistical and systematic components, reaching sub-GeV precision for the first time in this event topology. The masses of the top quark and antiquark are also determined separately using the lepton charge in the final state, from which the mass ratio and difference are determined to be 0.9952$^{+0.0079}_{-0.0104}$ and 0.83$^{+1.79}_{-1.35}$ GeV, respectively. The results are consistent with $CPT$ invariance.

Measurements of the inclusive and differential fiducial cross sections of the Higgs boson are presented, using the $\tau$ lepton decay channel. The differential cross sections are measured as functions of the Higgs boson transverse momentum, jet multiplicity, and transverse momentum of the leading jet in the event if any. The analysis is performed using proton-proton data collected with the CMS detector at the LHC at a center-of-mass energy of 13 TeV and corresponding to an integrated luminosity of 138 fb$^{-1}$. These are the first differential measurements of the Higgs boson cross section in the final state of two $\tau$ leptons, and they constitute a significant improvement over measurements in other final states in events with a large jet multiplicity or with a Lorentz-boosted Higgs boson.

The associated production of a W and a Z boson is studied in final states with multiple leptons produced in proton-proton (pp) collisions at a centre-of-mass energy of 13 TeV using 137 fb$^{-1}$ of data collected with the CMS detector at the LHC. A measurement of the total inclusive production cross section yields $\sigma_{\text{tot}}$(pp $\to$ WZ) = 50.6 $\pm$ 0.8 (stat) $\pm$ 1.5 (syst) $\pm$ 1.1 (lum) $\pm$ 0.5 (thy) pb. Measurements of the fiducial and differential cross sections for several key observables are also performed in all the final-state lepton flavour and charge compositions with a total of three charged leptons, which can be electrons or muons. All results are compared with theoretical predictions computed up to next-to-next-to-leading order in quantum chromodynamics plus next-to-leading order in electroweak theory and for various sets of parton distribution functions. The results include direct measurements of the charge asymmetry and the W and Z vector boson polarization. The first observation of longitudinally polarized W bosons in WZ production is reported. Anomalous gauge couplings are searched for, leading to new constraints on beyond-the-standard-model contributions to the WZ triple gauge coupling.

A search for new long-lived particles decaying to leptons using proton-proton collision data produced by the CERN LHC at $\sqrt{s}$ = 13 TeV is presented. Events are selected with two leptons (an electron and a muon, two electrons, or two muons) that both have transverse impact parameter values between 0.01 and 10 cm and are not required to form a common vertex. Data used for the analysis were collected with the CMS detector in 2016, 2017, and 2018, and correspond to an integrated luminosity of 118 (113) fb$^{-1}$ in the ee channel (e$\mu$ and $\mu\mu$ channels). The search is designed to be sensitive to a wide range of models with displaced e$\mu$, ee, and $\mu\mu$ final states. The results constrain several well-motivated models involving new long-lived particles that decay to displaced leptons. For some areas of the available phase space, these are the most stringent constraints to date.

The CMS experiment at the LHC has measured the differential cross sections of Z bosons decaying to pairs of leptons, as functions of transverse momentum and rapidity, in lead-lead collisions at a nucleon-nucleon center-of-mass energy of 5.02 TeV. The measured Z boson elliptic azimuthal anisotropy coefficient is compatible with zero, showing that Z bosons do not experience significant final-state interactions in the medium produced in the collision. Yields of Z bosons are compared to Glauber model predictions and are found to deviate from these expectations in peripheral collisions, indicating the presence of initial collision geometry and centrality selection effects. The precision of the measurement allows, for the first time, for a data-driven determination of the nucleon-nucleon integrated luminosity as a function of lead-lead centrality, thereby eliminating the need for its estimation based on a Glauber model.

Measurements of differential and double-differential cross sections of top quark pair ($\text{t}\overline{\text{t}}$) production are presented in the lepton+jets channels with a single electron or muon and jets in the final state. The analysis combines for the first time signatures of top quarks with low transverse momentum $p_\text{T}$, where the top quark decay products can be identified as separated jets and isolated leptons, and with high $p_\text{T}$, where the decay products are collimated and overlap. The measurements are based on proton-proton collision data at $\sqrt{s} = $ 13 TeV collected by the CMS experiment at the LHC, corresponding to an integrated luminosity of 137 fb$^{-1}$. The cross sections are presented at the parton and particle levels, where the latter minimizes extrapolations based on theoretical assumptions. Most of the measured differential cross sections are well described by standard model predictions with the exception of some double-differential distributions. The inclusive $\text{t}\overline{\text{t}}$ production cross section is measured to be $\sigma_{\text{t}\overline{\text{t}}} = $ 791 $\pm$ 25 pb, which constitutes the most precise measurement in the lepton+jets channel to date.