arXiv subjects

Henning Bahl

Publications and source records attributed to Henning Bahl.

At least 19 recordsLinked to original sources

arXiv2026

Generative Amplification with Surrogate Monte Carlo

Amplitude surrogates for LHC simulations build on generative amplification, the fact that a surrogate trained on an expensive and small training dataset describes the smooth amplitude more precisely than the training data does. Applying techniques developed for generative networks, we quantify this amplification for gluon-associated $Z$ production. Significant amplification appears in sparsely populated kinematic tails, where it matters most. Our results show how generative amplification from surrogate Monte Carlo far outperforms the density estimation in current generative networks.

hep-ph↗

arXiv2026

HiggsTools for LHC Run 3 and Beyond

HiggsTools, including the subpackages HiggsPredictions, HiggsBounds, and HiggsSignals, is a toolbox for Beyond-the-SM (BSM) scalar phenomenology at the LHC. It provides BSM model predictions, tests the model against experimental limits from searches for BSM scalars and derives constraints from the measurements of the properties of the discovered Higgs boson. We present a variety of improvements to the HiggsTools framework, preparing it for the results of LHC Run 3 and the HL-LHC. HiggsPredictions now provides additional cross-section predictions for centre-of-mass energies of 13.6 and 14 TeV. Moreover, it now includes cross-section predictions for resonant and non-resonant Higgs-boson pair production. For HiggsBounds, we describe the recasting of searches using multi-top final states, explain their implementation and highlight the impact of the experimental sensitivity of those results. Furthermore, we discuss the implementation of coupling-dependent limits on non-resonant Higgs boson pair production, as well as the improved handling of searches conducted prior to the Higgs boson discovery. For HiggsSignals we describe several improvements for the case of scalars with mass uncertainties.

hep-ph↗

arXiv2026

Local Conformal Predictions for Calibrated Surrogates

Neural network surrogates for LHC scattering amplitudes require trustworthy uncertainty estimates, a challenging task given the non-Gaussian systematics. We target it using conformal prediction, a distribution-free post-processing to complement trained surrogates with calibrated uncertainties. We find that standard conformal predictions struggle to provide locally calibrated uncertainties. This leads us to introduce FALCON, a novel conformal prediction method that learns locally calibrated confidence intervals. Our simple examples illustrate the power of distribution-free uncertainty quantification for ultra-fast event generation at the LHC.

hep-ph↗

arXiv2026

One Generator, Any Process: LLM-Conditioning for the LHC

Neural network training for LHC event generation should, ideally, benefit from common high-level patterns in different processes. We propose novel conditioning schemes for continuous parameters, process labels, and Feynman diagrams. We employ pre-trained LLMs as multi-modal foundation models to provide descriptive embeddings for an autoregressive transformer. With such high-level physics-inductive bias the generative networks converge faster, provide better result, and generalize to unseen processes.

hep-ph↗

arXiv2026

Precise predictions for trilinear Higgs couplings and Higgs pair production in extended scalar sectors with anyH3 and anyHH

A central objective of future collider experiments is to probe the structure of the Higgs potential, which requires access to trilinear scalar couplings, in particular the self-coupling of the observed Higgs boson. While this coupling is fixed in the Standard Model (SM), it can receive sizable modifications in many Beyond the SM (BSM) scenarios, often connected to solutions of open problems such as the origin of the matter-antimatter asymmetry of the Universe. In theories with extended scalar sectors, radiative corrections involving additional scalar states can significantly affect both the Higgs self-coupling and other trilinear scalar interactions, with important consequences for predictions of physical observables. Precise theoretical calculations are therefore essential for the interpretation of precision Higgs measurements and for identifying indirect signatures of new physics. This contribution presents the latest version of the public tool anyBSM, which provides automated calculations of all trilinear scalar couplings at full one-loop order in arbitrary renormalisable theories, including full momentum dependence and flexible renormalisation-scheme choices. In addition, the new module anyHH for di-Higgs production in gluon fusion is discussed in several exemplary BSM models, including scenarios with multiple resonances.

hep-ph↗

arXiv2026

Sensitivity to new physics: single-Higgs couplings vs. the trilinear Higgs coupling

The trilinear Higgs self-coupling provides a unique probe of the structure of the Higgs potential and of the nature of the electroweak phase transition, and constitutes a key target for future collider experiments. Recent studies have shown that confronting theoretical predictions for the trilinear Higgs coupling with current experimental bounds offers a powerful and complementary way to test effects of physics beyond the Standard Model (BSM), in particular those arising from extended Higgs sectors. Meanwhile, substantial progress has been achieved in the precise calculation and automation of the trilinear Higgs coupling in a wide class of BSM models. This contribution discusses several BSM scenarios, compatible with existing constraints, in which sizeable deviations in the trilinear Higgs coupling w.r.t. the Standard Model (SM) value are predicted, while other Higgs observables remain close to their SM expectations and are therefore difficult to probe experimentally. These results highlight the strong physics motivation for a precise measurement of the trilinear Higgs coupling at a future Higgs factory.

hep-ph↗

arXiv2026

Complementarity of di-top and four-top searches in interpreting possible signals of new physics

Final states comprising two or more top quarks are important search channels at the Large Hadron Collider for scalar particles predicted in models of physics beyond the Standard Model. While the di-top final state profits from a higher signal cross section, it can be subject to intricate interference patterns. Besides the interference with the large QCD background, in case of the presence of more than one high-mass scalar also large signal--signal interference contributions can occur. We show that in such scenarios it is crucial to account for loop-level mixing for obtaining accurate exclusion bounds. We demonstrate how the interference patterns can obscure the interpretation of possible deviations from the Standard Model expectations. We show that the four-top final state, while giving rise to a smaller signal cross section, provides important complementary information due to its much smaller signal--background interference contributions. Thus, the results obtained from the four-top final state can be instrumental for pinpointing the underlying new physics scenario.

hep-ph↗

arXiv2026

Scaling laws for amplitude surrogates

Scaling laws describing the dependence of neural network performance on the amount of training data, the spent compute, and the network size have emerged across a huge variety of machine learning task and datasets. In this work, we systematically investigate these scaling laws in the context of amplitude surrogates for particle physics. We show that the scaling coefficients are connected to the number of external particles of the process. Our results demonstrate that scaling laws are a useful tool to achieve desired precision targets.

hep-ph↗

arXiv2026

How to Trust Learned Loop Amplitudes

Higher-order theory predictions are crucial for the precision LHC program, but the time-consuming amplitude evaluation challenges the corresponding Monte-Carlo simulations. Machine-learned amplitude surrogates can resolve this problem, if we can guarantee their precision over the entire phase space. First, we show that our surrogates provide a calibrated learned uncertainty, even for non-Gaussian systematics; second, we describe how less accurate phase space regions can be identified; third, we demonstrate how the precision in these regions can be improved reliably.

hep-ph↗

arXiv2025

Assessing uncertainties in the determination of the trilinear Higgs self-coupling from single-Higgs observables

Circular $e^{+}e^{-}$ colliders operating at energies below the di-Higgs production threshold can provide information on the trilinear Higgs self-coupling $\lambda_{hhh}$ via its loop contributions to single Higgs production processes and electroweak precision observables. We investigate how well a non-SM value of $\lambda_{hhh}$ can be determined indirectly via its loop contributions within a global EFT fit. Using an inert doublet extension of the SM Higgs sector as an example for a scenario of physics beyond the SM that could be realised in nature, we find that theoretical uncertainties related to the treatment of loop contributions and the truncation of the EFT expansion, which are usually neglected, play an important role in determining the sensitivity to $\lambda_{hhh}$ in a global fit. The results obtained from such an indirect determination of $\lambda_{hhh}$ without taking these additional uncertainties into account would be too optimistic, leading to an artificially high resulting precision for $\lambda_{hhh}$. They could therefore be misleading in the quest to precisely identify the underlying physics of electroweak symmetry breaking.

hep-ph↗

arXiv2025

Forecasting Generative Amplification

Generative networks are perfect tools to enhance the speed and precision of LHC simulations. Especially when generating events beyond the size of the training dataset, it is important to understand their statistical precision. We present two complementary methods to estimate the amplification factor without large holdout datasets. Averaging amplification uses Bayesian networks or ensembling to estimate amplification from the precision of integrals over given phase-space volumes. Differential amplification uses hypothesis testing to quantify amplification without any resolution loss. Applied to state-of-the-art event generators, both methods indicate that amplification is already possible in specific regions of phase space.

hep-ph↗

arXiv2025

Unbinning global LHC analyses

Neural simulation-based inference has been shown to outperform traditional, histogram-based inference in numerous phenomenological and experimental studies at the LHC. So far, these analyses have focused on individual processes. We study the combination of four different di-boson processes in terms of the Standard Model Effective Field Theory. Our results demonstrate how neural simulation-based inference also wins over traditional methods for more global LHC analyses.

hep-ph↗

arXiv2025

Amplitude Uncertainties Everywhere All at Once

Ultra-fast, precise, and controlled amplitude surrogates are essential for future LHC event generation. First, we investigate the noise reduction and biases of network ensembles and outline a new method to learn well-calibrated systematic uncertainties for them. We also establish evidential regression as a sampling-free method for uncertainty quantification. In a second part, we tackle localized disturbances for amplitude regression and demonstrate that learned uncertainties from Bayesian networks, ensembles, and evidential regression all identify numerical noise or gaps in the training data.

hep-ph↗

arXiv2025

$\mathcal{CP}$-Analyses with Symbolic Regression

Searching for $\mathcal{CP}$ violation in Higgs interactions at the LHC is as challenging as it is important. Although modern machine learning outperforms traditional methods, its results are difficult to control and interpret, which is especially important if an unambiguous probe of a fundamental symmetry is required. We propose solving this problem by learning analytic formulas with symbolic regression. Using the complementary PySR and SymbolNet approaches, we learn $\mathcal{CP}$-sensitive observables at the detector level for WBF Higgs production and top-associated Higgs production. We find that they offer advantages in interpretability and performance.

hep-ph↗

arXiv2025

Generic two-loop results for trilinear and quartic scalar self-interactions

Reconstructing the shape of the Higgs potential realised in Nature is a central part of the physics programme at the LHC and future colliders. In this context, accurate theoretical predictions for trilinear and quartic Higgs couplings are becoming increasingly important. In this paper, we present results that enable significant progress in the automation of these calculations at the two-loop level in a wide range of models. Specifically, we calculate the generic two-loop corrections for scalar n-point functions with n<=4 assuming that all external scalars are identical. Working in the zero-momentum approximation, we express the results in terms of generic couplings and masses. Additionally, by exploiting permutation invariances, we reduce the number of Feynman diagrams appearing to a substantially smaller set of basis diagrams. To ease the application of our setup, we also provide routines that allow to map our generic results to scalar two-loop amplitudes generated with the package FeynArts. We perform a series of calculations to cross-check our results with existing results in the literature. Moreover, we present new two-loop results for the trilinear Higgs coupling in the general singlet extension of the Standard Model. We also present the public Python package Tintegrals, which allows for fast and stable evaluations of all relevant two-loop integrals with vanishing external momenta.

hep-ph↗

arXiv2025

Impact of Interference Effects on Higgs-boson Searches in the Di-top Final State at the LHC

The di-top final state is an important search channel for additional Higgs bosons at the LHC. In this channel, large signal--background interference contributions can strongly distort a resonance peak as it would be expected from a pure signal contribution. Moreover, signal--signal interference effects can have a significant impact if more than one additional scalar particle is present. In this work, we perform a comprehensive model-independent analysis of the various interference contributions considering two additional heavy scalars that can mix with each other. We point out the importance of taking into account loop-level mixing between the scalars. A proper treatment of these mixing effects, which has not been previously carried out for the di-top final state, introduces additional relative phases between different parts of the amplitudes entering the interference contributions which we find to have a strong impact on the di-top invariant mass distribution. We study the interference effects both in an idealistic setting as well as taking into account experimental limitations using Monte-Carlo simulations. We demonstrate that the emerging experimental signatures can be unexpected and difficult to interpret. In particular, we point out that an experimental signature manifesting itself as an excess near the $t \bar t$ threshold may actually be caused by new scalar particles with much higher masses. We comment in this context on the recent excess that has been observed by the CMS collaboration near the $t \bar t$ threshold in their searches in the di-top final state.

hep-ph↗

arXiv2024

Accurate Surrogate Amplitudes with Calibrated Uncertainties

Neural networks for LHC physics have to be accurate, reliable, and controlled. Using neural surrogates for the prediction of loop amplitudes as a use case, we first show how activation functions are systematically tested with Kolmogorov-Arnold Networks. Then, we train neural surrogates to simultaneously predict the target amplitude and an uncertainty for the prediction. We disentangle systematic uncertainties, learned by a well-defined likelihood loss, from statistical uncertainties, which require the introduction of Bayesian neural networks or repulsive ensembles. We test the coverage of the learned uncertainties using pull distributions to quantify the calibration of cutting-edge neural surrogates.

hep-ph↗

arXiv2024

Topportunities at the LHC: Rare Top Decays with Light Singlets

The discovery of the top quark, the most massive elementary particle yet known, has given us a distinct window into investigating the physics of the Standard Model and Beyond. With a plethora of top quarks to be produced in the High Luminosity era of the LHC, the exploration of its rare decays holds great promise in revealing potential new physics phenomena. We consider higher-dimensional operators contributing to top decays in the SMEFT and its extension by a light singlet species of spin 0, 1/2, or 1, and exhibit that the HL-LHC may observe many exotic top decays in a variety of channels. Light singlets which primarily talk to the SM through such a top interaction may also lead to distinctive long-lived particle signals. Searching for such long-lived particles in top-quark decays has the additional advantage that the SM decay of the other top quark in the same event provides a natural trigger.

hep-ph↗