arXiv subjects

Veronica Sanz

Publications and source records attributed to Veronica Sanz.

At least 19 recordsLinked to original sources

arXiv2026

Language-Guided Hypotheses Generation for Sparse SMEFT Analyses

Global fits of the Standard Model Effective Field Theory are challenged by the large number of operators, while any given database constrains only a small subset. Selecting relevant operator hypotheses therefore requires theoretical insight into operator correlations and the sensitivity of observables. We present llm4smeft, an open source framework that addresses this problem by combining a language model, fine-tuned on the SMEFT literature, with retrieval augmented generation based on quantitative summaries of SMEFiT package global fits. Given a set of observables, the framework proposes candidate relevant operators together with their corresponding Fisher information, while retrieval ensures that model outputs are grounded in existing fit results whenever available. The framework runs in an interactive mode in which accepted hypotheses are stored in a growing knowledge base. We publicly release the llm4smeft package together with the fine-tuned language model, in which the entire framework runs locally, requiring neither internet access nor paid cloud services.

hep-ph↗

arXiv2026

IRIS-GAN: Staged Specialist Detection of Deepfake Faces

We introduce IRIS-GAN, a specialist forensic detector for synthetic face images under cross-generator shift. Rather than addressing universal synthetic-image detection, we focus on faces generated by generative adversarial networks (GANs), which are state-of-the-art in deepfake content, and train the detector through staged exposure to increasingly demanding GAN families while retaining earlier generators. The final model reaches fake-detection rates above 99% across the GAN families considered and classifies an external real-face dataset with 98.9% accuracy. Grad-CAM analysis further reveals measurable generator-dependent spatial response patterns, which remain informative for a secondary heatmap-only classifier. Out-of-family tests on diffusion-generated faces confirm that IRIS-GAN is a specialist detector, with some capability to reach non-GAN deepfakes. These results establish staged training as an effective strategy for robust GAN-face forensics.

cs.CV↗

arXiv2026

Operational tracking loss in nonautonomous second-order oscillator networks

We study when a network of coupled oscillators with inertia ceases to follow a time-dependent driving protocol coherently, using a simplified graph-based model motivated by inverter-dominated energy systems. We show that this loss of tracking is diagnosed most clearly in the frequency dynamics, rather than in phase-based observables. Concretely, a tracking ratio built from the frequency-disagreement observable $E_ω(t)$ and normalized by the instantaneous second-order modal decay rate yields a robust protocol-dependent freeze-out time whose relative dispersion decreases with system size. Graph topology matters substantially: the resulting freeze-out time is only partly captured by the algebraic connectivity $λ_2$, while additional structural descriptors, particularly Fiedler-mode localization and low-spectrum structure, improve the explanation of graph-to-graph variation. By contrast, phase-sector observables develop strong non-monotonic and underdamped structure, so simple diagonal low-mode relaxation closures are not quantitatively reliable in the same regime. These results identify the frequency sector as the natural operational sector for nonautonomous tracking loss in second-order oscillator networks and clarify both the usefulness and the limits of reduced spectral descriptions in this setting.

nlin.AO↗

arXiv2026

Adversarial Stress Tests for Quantum Certification

We develop a practical framework for semi-device-independent (SDI) certification under operational deviations from the ideal protocol model. Apparent violations of classical benchmarks need not signal genuinely non-classical behaviour; they can arise from misalignment between (i) the scoring rule, (ii) the finite-sample statistical bound applied to that score, and (iii) the operational model realised in the experiment, including bias, memory, drift, and selection effects. We formalise a protocol-agnostic alignment principle based on a martingale-safe lower confidence bound and an operationally consistent effective classical ceiling. This yields a quantitative diagnostic, the \emph{robustness gap} $Δ_{\mathrm{rob}} = S_{\mathrm{low}} - S_{C,\mathrm{eff}}$, which separates statistical fluctuations from structural modelling errors. Statistical deviations vanish asymptotically, whereas model misalignment can produce persistent false certification unless the benchmark is corrected. Using the $2\!\to\!1$ random access code as a minimal SDI testbed, we show that postselection can inflate conditional scores, whereas unconditional scoring restores the correct operational meaning of the witness. We further show that adaptive learning-based classical agents do not enlarge the admissible classical set; rather, they recover the effective classical ceiling implied by the operational model. The resulting framework provides a systematic diagnostic for certification in realistic quantum communication and measurement settings with embedded classical control, adaptive processing, and nonideal data acquisition.

quant-ph↗

arXiv2026

Machine-Learning-Inspired SMEFT Simplified Template Cross Sections: A Case Study in ZH Production

The Simplified Template Cross Section (STXS) program has become the standard interface between Higgs measurements and global fits, but its fixed one-dimensional boundaries are not guaranteed to align with the phase-space directions to which the Standard Model Effective Field Theory (SMEFT) is most sensitive. We propose a machine-learning-inspired extension of STXS in which supervised classifiers are used only at the design stage to identify simple, publishable phase-space boundaries. Using associated Higgs production, $pp \to ZH$, as a case study and a benchmark momentum-dependent bosonic SMEFT deformation, we show that the relevant signal-background separation is well captured by a linear boundary in the $(p_T^Z,mZH)$ plane. We construct such boundaries with a linear support vector machine and with a deep-neural-network-assisted distillation procedure, and compare them directly with the standard STXS $p_T^Z$ bins through a common single-region Asimov-significance analysis. In this proof-of-concept setup, the ML-inspired regions systematically outperform the corresponding STXS regions, with the largest gains appearing in the boosted regime where SMEFT effects are concentrated. The final observable remains a simple linear cut, preserving the transparency and experimental portability that make STXS useful.

hep-ph↗

arXiv2026

Operational Emergence of a Global Phase under Time-Dependent Coupling in Oscillator Networks

Time-dependent coupling is often interpreted as introducing competition between a protocol rate and an intrinsic synchronization rate. We show that this interpretation is unavailable in the identical, overdamped Kuramoto model when time dependence enters only through a scalar multiplier of a fixed coupling field. The accumulated coupling $S(t)=\int_0^tK(u)\mathrm{d} u$ is then an exact dynamical clock: equal-exposure protocols traverse the same autonomous orbit, and integrable decays produce finite-exposure arrest rather than deterministic loss of adiabatic tracking. This clock suggests a classification under time reparametrization. Deterministic orbit observables and winding sectors are invariant; an additive perturbation $\varepsilon G$ has a leading response that is a linear functional of the reciprocal schedule $w(s)=1/K(t(s))$ against a kernel fixed by the autonomous orbit; and inertia or noncommuting graph generators lie outside this reciprocal-weighted class. Laboratory-time white noise has the corresponding $w$-weighted covariance. We derive exact mode covariances, prove a universal terminal bang--bang optimum under bounded coupling and fixed exposure, and validate both results in nonlinear networks. We also separate information in the full phase sample from information retained by the order parameter. Conditional observation error of $\arg Z$ follows a $1/(NR^2)$ law with a configuration-dependent prefactor, whereas a known-template experiment has Fisher information $N/\sigma^2$ independently of $R$; the efficiency of $\arg Z$ is $R^2/q_2$. Finally, periodic rings exhibit stable winding sectors with strong local but vanishing global order.

cond-mat.stat-mech↗

arXiv2026

Artificial Intelligence and Symmetries: Learning, Encoding, and Discovering Structure in Physical Data

Symmetries play a central role in physics, organizing dynamics, constraining interactions, and determining the effective number of physical degrees of freedom. In parallel, modern artificial intelligence methods have demonstrated a remarkable ability to extract low-dimensional structure from high-dimensional data through representation learning. This review examines the interplay between these two perspectives, focusing on the extent to which symmetry-induced constraints can be identified, encoded, or diagnosed using machine learning techniques. Rather than emphasizing architectures that enforce known symmetries by construction, we concentrate on data-driven approaches and latent representation learning, with particular attention to variational autoencoders. We discuss how symmetries and conservation laws reduce the intrinsic dimensionality of physical datasets, and how this reduction may manifest itself through self-organization of latent spaces in generative models trained to balance reconstruction and compression. We review recent results, including case studies from simple geometric systems and particle physics processes, and analyze the theoretical and practical limitations of inferring symmetry structure without explicit inductive bias.

hep-ph↗

arXiv2026

Was the Early Universe Quantum? Falsifying Classical Stochastic Inflation

Inflationary cosmology successfully accounts for the observed properties of primordial fluctuations using quantum field theory in an expanding background. However, the quantum nature of these fluctuations has not been experimentally established, since classical stochastic models could reproduce the observed two-point statistics by construction. Existing approaches to testing primordial quantumness focus primarily on Bell inequalities, which provide a sharp conceptual criterion but are difficult to implement with cosmological observables. In this work we adopt a falsification-based approach. We define a precise classical hypothesis for the origin of primordial perturbations (local stochastic fields admitting a positive probability distribution) and identify inequality constraints that must be satisfied within this class. We show how violations of these classicality inequalities can be probed using realistic cosmological observables, without invoking Bell tests or non-commuting measurement settings. We further identify symmetry-protected spectator sectors in which quantum coherence is parametrically preserved during inflation, allowing violations of observable magnitude to survive decoherence. Our results show that large-scale structure and future 21 cm surveys provide a viable and quantitative route to falsifying classical stochastic descriptions of primordial fluctuations.

astro-ph.CO↗

arXiv2026

Axion misalignment as a synchronization phenomenon

We propose a dynamical reinterpretation of axion misalignment as an emergent collective phenomenon. Drawing an explicit parallel between axion field dynamics and synchronization in coupled oscillator systems, we show that a macroscopic axion phase can arise dynamically from initially incoherent configurations through gradient-driven ordering in an expanding Universe. In this framework, the misalignment angle is not a fundamental initial condition but a collective variable that becomes well defined only once phase coherence develops. Using a three-dimensional lattice study with full second-order dynamics, we show that this collective phase is selected prior to the onset of axion oscillations in the regimes studied, providing dynamical support for the standard misalignment picture. This perspective offers a new way of organizing axion initial-condition sensitivity and reframes small-angle assumptions in terms of the efficiency of phase ordering in the early Universe.

astro-ph.CO↗

arXiv2025

Angular Coefficients from Interpretable Machine Learning with Symbolic Regression

We explore the use of symbolic regression to derive compact analytical expressions for angular observables relevant to electroweak boson production at the Large Hadron Collider (LHC). Focusing on the angular coefficients that govern the decay distributions of $W$ and $Z$ bosons, we investigate whether symbolic models can well approximate these quantities, typically computed via computationally costly numerical procedures, with high fidelity and interpretability. Using the PySR package, we first validate the approach in controlled settings, namely in angular distributions in lepton-lepton collisions in QED and in leading-order Drell-Yan production at the LHC. We then apply symbolic regression to extract closed-form expressions for the angular coefficients $A_i$ as functions of transverse momentum, rapidity, and invariant mass, using next-to-leading order simulations of $pp \to \ell^+\ell^-$ events. Our results demonstrate that symbolic regression can produce accurate and generalisable expressions that match Monte Carlo predictions within uncertainties, while preserving interpretability and providing insight into the kinematic dependence of angular observables.

hep-ph↗

arXiv2025

Data-driven discovery strategy for standard model effective field theory searches

We present a novel strategy to uncover indirect signs of new physics in collider data using the Standard Model Effective Field Theory (SMEFT) framework, offering notably improved sensitivity compared to traditional global analyses. Our approach leverages genetic algorithms to efficiently navigate the high-dimensional space of operator subsets, identifying deformations that improve agreement with data without relying on prior UV assumptions. This enables the systematic detection of SMEFT scenarios that outperform the Standard Model in explaining observed deviations. We validate the approach on current LHC and LEP measurements, perform closure tests with injected UV signals, and assess performance under high-luminosity projections. The algorithm successfully recovers relevant operator subsets and highlights directions in parameter space where deviations are most likely to emerge. Our results demonstrate the potential of SMEFT-based discovery searches driven by model selection, providing a scalable framework for future data analyses.

hep-ph↗

arXiv2025

A global analysis of ALP-mediated multiboson production at the LHC

Axion-like particles (ALPs) provide a well-motivated framework for physics beyond the Standard Model, coupling to gauge bosons through dimension-five operators protected by an approximate shift symmetry. At the LHC, such interactions lead to distinctive signatures in multiboson production, where the ALP appears as an off-shell mediator rather than a narrow resonance. In this work, we present the first global analysis of ALP-mediated multiboson processes, combining measurements of diphoton, ZZ, $W^+ W^-$, dijet, and vector-boson-fusion final states. On the theory side, motivated from a UV perspective, we assume that the ALP couples only to the gauge sector of the SM, and classify the ALP-multiboson vertices that directly govern collider observables. Our results show that the dijet channel dominates the sensitivity to ALP couplings and determines the limits on $c_{\tilde{G}}$, while diboson and VBF processes provide complementary constraints on the electroweak couplings. We further assess the validity of the EFT expansion given the multi-TeV scales probed in the data. This global study provides the most comprehensive picture to date of ALP-gauge interactions from multiboson production at the LHC, and highlights the opportunities for significant improvements with future high-luminosity datasets.

hep-ph↗

arXiv2025

Probing the coupling of axions to tops and gluons with LHC measurements

We study axion-like particles (ALPs) whose dominant interactions are with gluons and third-generation quarks, and whose couplings to light Standard Model (SM) particles arise at one loop. These loop-induced effects lead to ALP decays and production channels that can be probed at the LHC, even when tree-level couplings are absent. Using an effective field theory (EFT) description that includes momentum-dependent corrections from radiative effects, we reinterpret a wide range of LHC measurements via the CONTUR framework to derive model-independent constraints on the ALP parameter space. We show that LHC data place meaningful bounds in the plane of effective couplings $c^0_t/f_a$ and $c^0_{\tilde G}/f_a$, and that these limits are sensitive to the UV origin of the ALP-top and ALP-gluon couplings. We discuss representative scenarios where either $c^0_t$ or $c^0_{\tilde G}$ vanishes at the matching scale, and highlight the role of EFT running and mixing in generating observable signals. We also assess the domain of validity of the EFT approach by comparing the typical momentum transfer $\sqrt{\hat s}$ in sensitive regions to the underlying scale $f_a$. Our results demonstrate the power of loop-aware EFT reinterpretation of SM measurements in probing otherwise elusive ALP scenarios. The framework presented here can be readily extended to include couplings to other fermions and to accommodate ALP decay or long-lived signatures.

hep-ph↗

arXiv2025

Strengthening Anomaly Awareness

We present a refined version of the Anomaly Awareness framework for enhancing unsupervised anomaly detection. Our approach introduces minimal supervision into Variational Autoencoders (VAEs) through a two-stage training strategy: the model is first trained in an unsupervised manner on background data, and then fine-tuned using a small sample of labeled anomalies to encourage larger reconstruction errors for anomalous samples. We validate the method across diverse domains, including the MNIST dataset with synthetic anomalies, network intrusion data from the CICIDS benchmark, collider physics data from the LHCO2020 dataset, and simulated events from the Standard Model Effective Field Theory (SMEFT). The latter provides a realistic example of subtle kinematic deviations in Higgs boson production. In all cases, the model demonstrates improved sensitivity to unseen anomalies, achieving better separation between normal and anomalous samples. These results indicate that even limited anomaly information, when incorporated through targeted fine-tuning, can substantially improve the generalization and performance of unsupervised models for anomaly detection.

hep-ph↗

arXiv2025

Learning symmetries in datasets

We investigate how symmetries present in datasets affect the structure of the latent space learned by Variational Autoencoders (VAEs). By training VAEs on data originating from simple mechanical systems and particle collisions, we analyze the organization of the latent space through a relevance measure that identifies the most meaningful latent directions. We show that when symmetries or approximate symmetries are present, the VAE self-organizes its latent space, effectively compressing the data along a reduced number of latent variables. This behavior captures the intrinsic dimensionality determined by the symmetry constraints and reveals hidden relations among the features. Furthermore, we provide a theoretical analysis of a simple toy model, demonstrating how, under idealized conditions, the latent space aligns with the symmetry directions of the data manifold. We illustrate these findings with examples ranging from two-dimensional datasets with $O(2)$ symmetry to realistic datasets from electron-positron and proton-proton collisions. Our results highlight the potential of unsupervised generative models to expose underlying structures in data and offer a novel approach to symmetry discovery without explicit supervision.

cs.LG↗

arXiv2025

Probing the flavour-blind SMEFT: EFT validity and the interplay of energy scales

The Standard Model Effective Field Theory (SMEFT) offers a systematic approach to study potential deviations from the Standard Model (SM) through higher-dimensional operators that encapsulate new physics effects. In this work, we analyze flavour-blind SMEFT contributions to flavour observables and assess their interplay with high-energy measurements from LEP and LHC. We perform global fits combining LEP precision data, flavour observables from rare B-meson decays, and LHC diboson measurements, revealing how the inclusion of different datasets breaks parameter degeneracies and enhances the sensitivity to SMEFT coefficients. Our study demonstrates that low-energy flavour observables provide reliable constraints even in flavour-blind scenarios, while high-energy measurements can be subject to EFT validity concerns due to kinematic growth. We investigate the impact of renormalization group evolution (RGE) and operator mixing across energy scales, highlighting the complementary nature of low- and high-energy datasets. The results emphasize the importance of flavour observables as robust probes of new physics and underline the necessity of global fits to avoid potential biases from limited datasets. Finally, we discuss the implications of our findings for the interpretation of global SMEFT analyses based on high-energy collider data, comparing UV models that contribute to SMEFT at tree- and loop-level.

hep-ph↗

arXiv2024

Symbolic regression for precision LHC physics

We study the potential of symbolic regression (SR) to derive compact and precise analytic expressions that can improve the accuracy and simplicity of phenomenological analyses at the Large Hadron Collider (LHC). As a benchmark, we apply SR to equation recovery in quantum electrodynamics (QED), where established analytical results from quantum field theory provide a reliable framework for evaluation. This benchmark serves to validate the performance and reliability of SR before extending its application to structure functions in the Drell-Yan process mediated by virtual photons, which lack analytic representations from first principles. By combining the simplicity of analytic expressions with the predictive power of machine learning techniques, SR offers a useful tool for facilitating phenomenological analyses in high energy physics.

hep-ph↗

arXiv2024

Gravity-Mediated Dark Matter at a low reheating temperature

We present a new study on the Gravity-Mediated Dark Matter (GMDM) scenario, where interactions between dark matter (DM) and the Standard Model are mediated by spin-two particles. Expanding on this established framework, we explore a novel regime characterized by a low reheating temperature that offers an alternative to the conventional thermal relic paradigm. This approach opens new possibilities for understanding the dynamics of the dark sector, encompassing both the dark matter particles (fermion, scalar and vector) and the spin-two mediators. Our analysis examines the constraints imposed by the relic abundance of DM, collider experiments, and direct detection searches, spanning a wide mass range for the dark sector, from very light to extremely heavy states. This work opens new possibilities for the phenomenology of GMDM.

hep-ph↗