Score: 1

Towards Trustworthy Amortized Bayesian Model Comparison

Published: August 28, 2025 | arXiv ID: 2508.20614v1

By: Šimon Kucharský , Aayush Mishra , Daniel Habermann and more

Potential Business Impact:

Helps computers pick the best explanation for data.

Business Areas:
A/B Testing Data and Analytics

Amortized Bayesian model comparison (BMC) enables fast probabilistic ranking of models via simulation-based training of neural surrogates. However, the reliability of neural surrogates deteriorates when simulation models are misspecified - the very case where model comparison is most needed. Thus, we supplement simulation-based training with a self-consistency (SC) loss on unlabeled real data to improve BMC estimates under empirical distribution shifts. Using a numerical experiment and two case studies with real data, we compare amortized evidence estimates with and without SC against analytic or bridge sampling benchmarks. SC improves calibration under model misspecification when having access to analytic likelihoods. However, it offers limited gains with neural surrogate likelihoods, making it most practical for trustworthy BMC when likelihoods are exact.

Country of Origin
🇩🇪 🇺🇸 Germany, United States

Page Count
13 pages

Category
Statistics:
Machine Learning (Stat)