Trustworthy scientific inference with generative models

Abstract

Generative artificial intelligence (AI) excels at producing complex data structures (text, images, videos) by learning patterns from training examples. Across scientific disciplines, researchers are now applying generative models to "inverse problems" to directly predict hidden parameters from observed data along with measures of uncertainty. While these predictive or posterior-based methods can handle intractable likelihoods and large-scale studies, they can also produce biased or overconfident conclusions even without model misspecifications. We present a solution with Frequentist-Bayes (FreB), a mathematically rigorous protocol that reshapes AI-generated posterior probability distributions into (locally valid) confidence regions that consistently include true parameters with the expected probability, while achieving minimum size when training and target data align. We demonstrate FreB's effectiveness by tackling diverse case studies in the physical sciences: identifying unknown sources under dataset shift, reconciling competing theoretical models, and mitigating selection bias and systematics in observational studies. By providing validity guarantees with interpretable diagnostics, FreB enables trustworthy scientific inference across fields where direct likelihood evaluation remains impossible or prohibitively expensive.

References in corpus (28)

• Gaia Data Release 3: Summary of the content and survey properties
• LSST: from Science Drivers to Reference Design and Anticipated Data Products
• Asymptotic formulae for likelihood-based tests of new physics
• The Apache Point Observatory Galactic Evolution Experiment (APOGEE)
• Approximate Bayesian computation scheme for parameter inference and model selection in dynamical systems
• A 3D Dust Map Based on Gaia, Pan-STARRS 1 and 2MASS
• Stellar Population Inference with Prospector
• Fermi-LAT Observations of Markarian 421: the Missing Piece of its Spectral Energy Distribution
• The Eighteenth Data Release of the Sloan Digital Sky Surveys: Targeting and First Spectra from SDSS-V
• Photo-astrometric distances, extinctions, and astrophysical parameters for Gaia EDR3 stars brighter than G=18.5
• Estimating stellar birth radii and the time evolution of the Milky Way's ISM metallicity gradient
• Least Ambiguous Set-Valued Classifiers with Bounded Error Levels
• Galactic Archaeology with Gaia
• Searching for new physics with profile likelihoods: Wilks and beyond
• Likelihood-Free Parameter Estimation with Neural Bayes Estimators
• OutbreakFlow: Model-based Bayesian inference of disease outbreak dynamics with invertible neural networks and its application to the COVID-19 pandemics in Germany
• SBI++: Flexible, Ultra-fast Likelihood-free Inference Customized for Astronomical Applications
• DESI Early Data Release Milky Way Survey Value-Added Catalogue
• Deciphering the evolution of the Milky Way discs: Gaia APOGEE Kepler giant stars and the Besançon Galaxy Model
• The quest to discover supersymmetry at the ATLAS experiment
• Scalable Spatiotemporal Prediction with Bayesian Neural Fields
• Approximate Bayesian Uncertainties on Deep Learning Dynamical Mass Estimates of Galaxy Clusters
• Component-wise approximate Bayesian computation via Gibbs-like steps
• An implementation of neural simulation-based inference for parameter estimation in ATLAS
• Frasian Inference
• Closing the stellar labels gap: Stellar label independent evidence for [$\alpha /M$] information in Gaia BP/RP spectra
• Mapping the Milky Way in 5-D with 170 Million Stars
• Six Maxims of Statistical Acumen for Astronomical Data Analysis