Abstract: In this talk, we introduce a novel methodology for addressing systematic uncertainties in unbinned inclusive cross-section measurements and related collider-based inference problems. Our approach incorporates known analytic dependencies on parameters of interest, including signal strengths and nuisance parameters. When these dependencies are unknown, as is frequently the case for systematic uncertainties, dedicated neural network parametrizations provide an approximation trained on simulated data. The resulting machine-learned surrogate captures the full parameter dependence of the likelihood ratio, providing a near-optimal test statistic. As a case study, we perform an inclusive cross-section measurement of H→𝜏𝜏 in the single-lepton channel, using simulated data from the FAIR Universe Higgs Uncertainty Challenge. Results on Asimov data, from large-scale toy studies, and based on Fisher information demonstrate significant improvements over traditional binned methods. In addition, I will very briefly introduce a complementary approach based on a graph neural network (GNN) that explicitly incorporates systematic variations into the training. Both the &Guaranteed Optimal “Log-Likelihood-based Unbinned Method” (GOLLUM) and the “Systematics-Aware Graph Estimator” (SAGE) codes are publicly available.

Abstract: Experiments at the Large Hadron Collider produce some of the most complex datasets in science. A central challenge in high-energy physics is to extract rare signals, reconstruct invisible particles such as neutrinos, and search for subtle deviations from the Standard Model with maximal sensitivity. These are fundamentally physics problems, but their scale and complexity call for new strategies. In this talk, I will present EveNet, a large-scale foundation model developed for collider physics. Trained on billions of simulated events, it provides a common starting point for diverse analysis tasks, including improving search sensitivity, reconstructing hidden structures, and detecting unexpected anomalies. By integrating such methods into the physicist’s toolkit, we show how foundation models can accelerate discovery and open new directions in the exploration of fundamental laws of nature.

Abstract: The practice of collider physics typically involves the marginalization of multi-dimensional collider data to uni-dimensional observables relevant for some physics task. In any cases, such as classification or anomaly detection, the observable can be arbitrarily complicated, such as the output of a neural network. However, for precision measurements, the observable must correspond to something computable systematically beyond the level of current simulation tools. In this work, we demonstrate that precision-theory-compatible observable space exploration can be systematized by using neural simulation-based inference techniques from machine learning. We illustrate this approach by exploring the space of marginalizations of the energy 3-point correlator to optimize sensitivity to the the top quark mass. We first learn the energy-weighted probability density from simulation, then search in the space of marginalizations for an optimal triangle shape. Although simulations and machine learning are used in the process of observable optimization, the output is an observable definition which can be then computed to high precision and compared directly to data without any memory of the computations which produced it.

Abstract: tba

Abstract: tba

Abstract: tba

Abstract: tba

Abstract: tba

Abstract: tba

Abstract: In recent years various supervised learning methods that disentangle aleatoric and epistemic uncertainty estimates based on second-order distributions have been proposed. In this talk we will see that these methods often fail to capture critical components of epistemic uncertainty, particularly due to the mostly neglected component of model bias, but also due to only capturing procedural- and not data variability. To show this, we make use of a more fine-grained taxonomy of epistemic uncertainty sources in machine learning models, and analyse how the generalized bias-variance can be used to derive expressions for the different uncertainty components. By using a simulation-based evaluation protocol which encompasses epistemic uncertainty due to both procedural- and data-driven uncertainty components, we illustrate that current methods rarely capture the full spectrum of epistemic uncertainty. Through theoretical insights and experimental results on synthetic and real-world data, we find that high model bias can lead to misleadingly low estimates of epistemic uncertainty, and common second-order uncertainty quantification methods for regression seem to blur bias-induced errors into aleatoric estimates, thereby underrepresenting epistemic uncertainty. Our findings underscore that meaningful aleatoric estimates are feasible only if all relevant sources of epistemic uncertainty are properly represented.

Abstract: tba