arXiv subjects

Michael Krämer

Publications and source records attributed to Michael Krämer.

At least 19 recordsLinked to original sources

arXiv2026

Neural Scaling Laws for Jet Generation

Recently observed empirical scaling laws describe the performance of foundation-type models as three independent key quantities -- dataset size, compute, and model parameters -- are modified. Extracting these scaling laws informs the training of large complex models for which the tuning of hyperparameters in traditional ways is not feasible. This work for the first time explores if scaling laws can also be observed for the task of particle jet generation -- both relevant as a pre-training objective for foundation models and as in-situ simulation by itself. We indeed replicate the key logarithmic scaling law behavior for model-size scaling. Beyond studying the next token prediction validation loss of the generative model, we also study the sliced Wasserstein distance of five physical quantities that are not immediately available to the model during training. Our study shows that this quantity is monotonically related to the next token prediction validation loss, meaning that this loss is indeed a good proxy for the physics performance. For the scaling with dataset size and compute, we observe substantially weaker scaling behavior of both the loss and the sliced Wasserstein distance. We analyze this behavior by introducing the concept of a learnable window, and argue that autoregressive next token prediction on jet constituents exhibits comparatively rapid saturation relative to language-model studies. We discuss possible origins of this behavior, including the stochastic nature of QCD radiation and differences between generative and supervised learning tasks in collider physics.

hep-ph↗

arXiv2026

UFO Sightings with Matchete: Connecting BSM Lagrangians with Event Generators

We present a tool for the automated generation of Universal Feynman Output (UFO) files, used as input by many event generators, for generic Beyond the Standard Model (BSM) theories and their low-energy Effective Field Theory (EFT) descriptions. Building on the Matchete package, which provides a convenient interface for defining BSM models and deriving their EFTs via matching, our extension obtains the Lagrangian in the broken phase of the electroweak symmetry, performs automatic gauge fixing, computes all Feynman rules, imposes flavor symmetries, and exports the model in the UFO format. This significantly streamlines BSM and EFT analyses, as we illustrate with an EFT validity study for a multi-channel leptoquark search. As an application, we provide a complete UFO implementation of the Standard Model EFT up to dimension eight using the $M_W$ electroweak input scheme.

hep-ph↗

arXiv2026

Domain Adaptation against Background Sculpting in Anomaly Detection at the LHC

Weakly supervised anomaly detection has been shown to be an effective tool for model-agnostic searches for new physics, especially in the context of resonance searches. However, correlations between the anomaly score and the resonant mass can distort the background distribution after selecting on the anomaly score, complicating background estimation from the sidebands. To mitigate this background sculpting, we propose a domain-adaptation-based decorrelation of the anomaly score from the resonant mass. We study this approach using the LHC Olympics R&D data set and several weakly supervised anomaly detection methods. We find that domain adaptation can substantially reduce background sculpting while largely preserving the anomaly detection performance. When correlations between the input features and the resonant mass degrade the original method's performance, domain adaptation can also recover sensitivity.

hep-ph↗

arXiv2026

Machine Learning is Good for Physics - and Vice Versa

Scientific AI is rapidly transforming fundamental physics research and challenging defining aspects of the fundamental physics methodology. We discuss opportunities and dangers of this transformation and find exciting benefits from a close interaction between AI and fundamental physics, provided that we remain aware of the scientific methodologies of the respective fields. For fundamental physics, we discuss two such aspects: statistical validation and a generalizing theory description, both with the goal of discovering new physics in vast datasets.

hep-ph↗

arXiv2026

Toward a Comprehensive Exploration of Flavored Dark Matter Models

We present a comprehensive framework for the study of flavored dark matter models, combining relic density calculations with direct and indirect detection limits, collider constraints, and a global analysis of flavor observables based on SMEFT matching and renormalization-group evolution. The framework applies to scalar or fermionic dark matter, including both self-conjugate and non-self-conjugate cases. As a proof of principle, we analyze two scenarios with Majorana dark matter coupling to right-handed charged leptons and to right-handed down-type quarks, assuming a thermal freeze-out. In the leptophilic case, flavor-violating decays such as $μ\to e γ$ dominate the constraints, while LHC searches still leave sizable parameter space. For quark couplings, direct detection bounds and meson mixing severely restrict the allowed couplings, favoring hierarchical flavor structures. The toolchain presented in this paper is publicly available on GitHub (https://github.com/lena-ra/Flavored-Dark-Matter).

hep-ph↗

arXiv2026

Weakly supervised machine learning for model-agnostic searches of new phenomena in the $γ$-ray sky

The $γ$-ray sky, as observed by the Fermi Large Area Telescope, contains a significant number of unassociated sources that may point to new astrophysical populations or more exotic phenomena. Machine-learning methods are widely used for source classification and searches for new physics, but most existing approaches rely on fully supervised training and therefore on explicit signal models. We explore weakly supervised classification as a less model-dependent strategy for analysing $γ$-ray source spectra. In a background-versus-mixture setup, classifiers are trained on samples with different signal admixtures rather than on fully labelled signal and background events. We study three representative scenarios: pulsar-active galactic nuclei separation as a controlled benchmark, the identification of dark-matter subhalos, and spectral irregularities induced by axion-photon oscillations. In each case we investigate the impact of signal fraction and sample composition on classification performance. Our results show that weak supervision can identify anomalous or signal-like subsets of data while reducing the reliance on detailed signal templates during training. In favourable cases, the method approaches the performance of fully supervised classifiers, while remaining applicable in situations where the signal model is uncertain or only partially specified. Weakly supervised learning therefore provides a complementary candidate-selection and anomaly-ranking strategy for $γ$-ray data analysis and searches for new phenomena.

astro-ph.HE↗

arXiv2026

Kitchen Sink Anomaly Detection

An enormous amount of R&D effort has resulted in many new resonant anomaly detection methods being proposed in recent years. However, the vast majority of previous R&D studies have suffered from two limitations: they have focused on a very small set of simulated signal benchmark models; and they have either used small sets of carefully crafted high-level jet substructure observables, which can be highly performant but are prone to model dependence, or the full collider event phase space, which is more agnostic but suffers from reduced sensitivity. In this work, we address both limitations: we formulate a number of new simulated signal benchmarks, which we make publicly available in a format fully compatible with the LHCO R&D benchmark; and we explore a high-level, yet highly agnostic, observable set consisting of Energy Flow Polynomials in addition to the usual subjettiness variables. We evaluate this "kitchen sink" observable set for both an idealized anomaly detector and the CWoLa hunting task, along with three baseline observable sets (the Baseline LHC Olympics set, subjettiness observables, and Energy Flow Polynomials). We find that our kitchen sink approach is the most sensitive to a broad range of signal types. Furthermore, we show that an attribute bagging variant, in which each ensemble member is trained on a random subset of substructure observables, yields comparable anomaly detection performance while significantly reducing training cost.

hep-ph↗

arXiv2026

Agents of Discovery

The substantial data volumes encountered in modern particle physics and other domains of fundamental physics research allow (and require) the use of increasingly complex data analysis tools and workflows. While the use of machine learning (ML) tools for data analysis has recently proliferated, these tools are typically special-purpose algorithms that rely, for example, on encoded physics knowledge to reach optimal performance. In this work, we investigate a new and orthogonal direction: Using recent progress in large language models (LLMs) to create a team of agents -- instances of LLMs with specific subtasks -- that jointly solve data analysis-based research problems in a way similar to how a human researcher might: by creating code to operate standard tools and libraries (including ML systems) and by building on results of previous iterations. If successful, such agent-based systems could be deployed to automate routine analysis components to counteract the increasing complexity of modern tool chains. To investigate the capabilities of current-generation commercial LLMs, we consider the task of anomaly detection via the publicly available and highly-studied LHC Olympics dataset. Several current models by OpenAI (GPT-4o, o4-mini, GPT-4.1, and GPT-5) are investigated and their stability tested. Overall, we observe the capacity of the agent-based system to solve this data analysis problem. The best agent-created solutions mirror the performance of human state-of-the-art results.

hep-ph↗

arXiv2026

How to pick the best anomaly detector?

Anomaly detection has the potential to discover new physics in unexplored regions of the data. However, choosing the best anomaly detector for a given data set in a model-agnostic way is an important challenge which has hitherto largely been neglected. In this paper, we introduce the data-driven ARGOS metric, which has a sound theoretical foundation and is empirically shown to robustly select the most sensitive anomaly detection model given the data. Focusing on weakly-supervised, classifier-based anomaly detection methods, we show that the ARGOS metric outperforms other model selection metrics previously used in the literature, in particular the binary cross-entropy loss. We explore several realistic applications, including hyperparameter tuning as well as architecture and feature selection, and in all cases we demonstrate that ARGOS is robust to the noisy conditions of anomaly detection.

hep-ph↗

arXiv2025

A strategy for optimal material identification in solar dark photon absorption

Dark photons with masses in the 1-100 eV range can be produced in the Sun and subsequently absorbed in terrestrial detectors, offering a promising avenue for probing hidden-sector physics beyond the Standard Model. In this work, we develop a theoretically grounded strategy to identify optimal detector materials for solar dark photon absorption. Our strategy builds on a material-independent upper limit on the absorption rate, which we derive from Kramers-Kronig relations applied separately to the longitudinal and transverse dark photon modes. We show how the optimal material properties depend on the dark photon mass relative to the detector's plasma frequency, identifying the conditions under which a detector can saturate the theoretical upper limit. We then assess the performance of commonly used detector materials in light of these criteria and comment on the prospects of metamaterials featuring tunable plasma frequencies. Our results provide a general and model-independent framework to effectively guide the design of next-generation experiments targeting solar dark photons.

hep-ph↗

arXiv2025

Aspen Open Jets: Unlocking LHC Data for Foundation Models in Particle Physics

Foundation models are deep learning models pre-trained on large amounts of data which are capable of generalizing to multiple datasets and/or downstream tasks. This work demonstrates how data collected by the CMS experiment at the Large Hadron Collider can be useful in pre-training foundation models for HEP. Specifically, we introduce the AspenOpenJets dataset, consisting of approximately 178M high $p_T$ jets derived from CMS 2016 Open Data. We show how pre-training the OmniJet-$α$ foundation model on AspenOpenJets improves performance on generative tasks with significant domain shift: generating boosted top and QCD jets from the simulated JetClass dataset. In addition to demonstrating the power of pre-training of a jet-based foundation model on actual proton-proton collision data, we provide the ML-ready derived AspenOpenJets dataset for further public use.

hep-ph↗

arXiv2025

SECRET: Stochasticity Emulator for Cosmic Ray Electrons

The spectrum of cosmic-ray electrons depends sensitively on the history and spatial distribution of nearby sources. Given our limited observational handle on cosmic-ray sources, any model remains necessarily probabilistic. Previously, predictions were performed in a Monte Carlo fashion, summing the contributions from individual, simulated sources to generate samples from the statistical ensemble of possible electron spectra. Such simulations need to be re-run if the cosmic-ray transport parameters (e.g. diffusion coefficient, maximum energy) are changed, rendering any parameter study computationally expensive. In addition, a proper statistical analysis of observations and comparison with such probabilistic models requires the joint probability distribution of the full spectrum instead of only samples. Note that parametrising this joint distribution is rendered difficult by the non-Gaussian statistics of the cosmic-ray fluxes. Here, we employ machine learning to compute the joint probability distribution of cosmic-ray electron fluxes. Specifically, we employ masked autoregressive density estimation (MADE) for a representation of the high-dimensional joint probability distribution. In a first step, we train the network on a Monte Carlo simulation for a fixed set of transport parameters, thus significantly accelerating the generation of samples. In a second step, we extend this setup to SECRET (Stochasticity Emulator for Cosmic Ray Electrons), allowing to reliably interpolate over the space of transport parameters. We make the MADE and SECRET codes available at https://git.rwth-aachen.de/pmertsch/secret .

astro-ph.HE↗

arXiv2025

From Kernels to Features: A Multi-Scale Adaptive Theory of Feature Learning

Feature learning in neural networks is crucial for their expressive power and inductive biases, motivating various theoretical approaches. Some approaches describe network behavior after training through a change in kernel scale from initialization, resulting in a generalization power comparable to a Gaussian process. Conversely, in other approaches training results in the adaptation of the kernel to the data, involving directional changes to the kernel. The relationship and respective strengths of these two views have so far remained unresolved. This work presents a theoretical framework of multi-scale adaptive feature learning bridging these two views. Using methods from statistical mechanics, we derive analytical expressions for network output statistics which are valid across scaling regimes and in the continuum between them. A systematic expansion of the network's probability distribution reveals that mean-field scaling requires only a saddle-point approximation, while standard scaling necessitates additional correction terms. Remarkably, we find across regimes that kernel adaptation can be reduced to an effective kernel rescaling when predicting the mean network output in the special case of a linear network. However, for linear and non-linear networks, the multi-scale adaptive approach captures directional feature learning effects, providing richer insights than what could be recovered from a rescaling of the kernel alone.

cond-mat.dis-nn↗

arXiv2025

Reinterpretation and preservation of data and analyses in HEP

Data from particle physics experiments are unique and are often the result of a very large investment of resources. Given the potential scientific impact of these data, which goes far beyond the immediate priorities of the experimental collaborations that obtain them, it is imperative that the collaborations and the wider particle physics community publish and preserve sufficient information to ensure that this impact can be realised, now and into the future. The information to be published and preserved includes the algorithms, statistical information, simulations and the recorded data. This publication and preservation requires significant resources, and should be a strategic priority with commensurate planning and resource allocation from the earliest stages of future facilities and experiments.

hep-ph↗

arXiv2025

The Fundamental Limit of Jet Tagging

Identifying the origin of high-energy hadronic jets ('jet tagging') has been a critical benchmark problem for machine learning in particle physics. Jets are ubiquitous at colliders and are complex objects that serve as prototypical examples of collections of particles to be categorized. Over the last decade, machine learning-based classifiers have replaced classical observables as the state of the art in jet tagging. Increasingly complex machine learning models are leading to increasingly more effective tagger performance. Our goal is to address the question of convergence -- are we getting close to the fundamental limit on jet tagging or is there still potential for computational, statistical, and physical insights for further improvements? We address this question using state-of-the-art generative models to create a realistic, synthetic dataset with a known jet tagging optimum. Various state-of-the-art taggers are deployed on this dataset, showing that there is a significant gap between their performance and the optimum. Our dataset and software are made public to provide a benchmark task for future developments in jet tagging and other areas of particle physics.

hep-ph↗

arXiv2025

Semi-visible jets, energy-based models, and self-supervision

We present DarkCLR, a novel framework for detecting semi-visible jets at the LHC. DarkCLR uses a self-supervised contrastive-learning approach to create observables that are approximately invariant under relevant transformations. We use background-enhanced data to create a sensitive representation and evaluate the representations using a CLR-inspired anomaly score and a normalized autoencoder as density estimators. Our results show a remarkable sensitivity for a wide range of semi-visible jets and are more robust than a supervised classifier trained on a specific signal.

hep-ph↗

arXiv2025

Large Physics Models: Towards a collaborative approach with Large Language Models and Foundation Models

This paper explores ideas and provides a potential roadmap for the development and evaluation of physics-specific large-scale AI models, which we call Large Physics Models (LPMs). These models, based on foundation models such as Large Language Models (LLMs) - trained on broad data - are tailored to address the demands of physics research. LPMs can function independently or as part of an integrated framework. This framework can incorporate specialized tools, including symbolic reasoning modules for mathematical manipulations, frameworks to analyse specific experimental and simulated data, and mechanisms for synthesizing theories and scientific literature. We begin by examining whether the physics community should actively develop and refine dedicated models, rather than relying solely on commercial LLMs. We then outline how LPMs can be realized through interdisciplinary collaboration among experts in physics, computer science, and philosophy of science. To integrate these models effectively, we identify three key pillars: Development, Evaluation, and Philosophical Reflection. Development focuses on constructing models capable of processing physics texts, mathematical formulations, and diverse physical data. Evaluation assesses accuracy and reliability by testing and benchmarking. Finally, Philosophical Reflection encompasses the analysis of broader implications of LLMs in physics, including their potential to generate new scientific understanding and what novel collaboration dynamics might arise in research. Inspired by the organizational structure of experimental collaborations in particle physics, we propose a similarly interdisciplinary and collaborative approach to building and refining Large Physics Models. This roadmap provides specific objectives, defines pathways to achieve them, and identifies challenges that must be addressed to realise physics-specific large scale AI models.

physics.data-an↗

arXiv2024

$\overline{\text{D}}$arkRayNet: Emulation of cosmic-ray antideuteron fluxes from dark matter

Cosmic-ray antimatter, particularly low-energy antideuterons, serves as a sensitive probe of dark matter annihilating in our Galaxy. We study this smoking-gun signature and explore its complementarity with indirect dark matter searches using cosmic-ray antiprotons. To this end, we develop the neural network emulator $\overline{\text{D}}$arkRayNet, enabling a fast prediction of propagated antideuteron energy spectra for a wide range of annihilation channels and their combinations. We revisit the Monte Carlo simulation of antideuteron coalescence and cosmic-ray propagation, allowing us to explore the uncertainties of both processes. In particular, we take into account uncertainties from the $Λ_b$ production rate and consider two distinctly different propagation models. Requiring consistency with cosmic-ray antiproton limits, we find that AMS-02 shows sensitivity to a few windows of dark matter masses only, most prominently below 20 GeV. This region can be probed independently by the upcoming GAPS experiment. The program package $\overline{\text{D}}$arkRayNet is available on GitHub.

hep-ph↗