Dataset: Differentiable Multi-scale Effective Field Theory Likelihoods for Beyond the Standard Model Phenomenology

This dataset accompanies the paper "Differentiable Multi-scale Effective Field Theory Likelihoods for Beyond the Standard Model Phenomenology" by Aleks Smolkovic and Peter Stangl (arXiv:2603.15801). The files contain posterior samples and constraint summaries from fits to flavor-physics and Drell-Yan data perfomed in two effective field theories: Weak Effective Theory (WET): b→sℓℓ fit Fit to b→sμμ and b→see data in the WCxf basis "flavio" at scale μ = 4.8 GeV, constraining six Wilson coefficients. bsll.hdf5 — 200 000 posterior samples (6 parameters) bsll_Gaussian.json — Gaussian approximation (central values, standard deviations, correlation matrix) bsll_marginalized.json / bsll_profiled.json — marginalized and profiled 68% CL intervals in WCxf data format Standard Model Effective Field Theory (SMEFT): Drell-Yan + flavor fits Global fits in the WCxf bases "Warsaw" and "Warsaw up" at scale μ = 1 TeV, constraining 374 Wilson coefficients. Results for two sets of observables are provided: DY + b→sνν (DY_bsnunu.*) DY + flavor (DY_flavor.*) For each set of observables: *.hdf5 — 2 000 000 posterior samples in the "Warsaw up" WCxf basis (374 parameters, real and imaginary parts of Wilson coefficients listed separately) *_marginalized.json / *_profiled.json — marginalized and profiled 68% CL intervals, "Warsaw"/"Warsaw up" WCxf basis, WCxf data format Notebook: load_and_process_samples.ipynb — example notebook showing how to load and process the samples provided in HDF5 format and use the results to validate the bounds provided in JSON format. Format notes All HDF5 files share the same structure: file.hdf5├── attributes: "eft", "basis", "scale"├── dataset: "parameters" [n_parameters × 2, str] (rows: coefficient_name, "R" or "I" indicating real and imaginary parts)└── dataset: "samples" [n_samples × n_parameters, float32] Each column i of the "samples" dataset contains draws for the Wilson coefficient identified in row i of the "parameters" dataset, with the "R"/"I" flag indicating real and imaginary parts.