A computationally-tractable measure of global sensitivity for sampling-based Bayesian inference
Abstract
Bayesian inference can often be sensitive to the choice of hyperparameters of the prior or likelihood, yet defining and quantifying this sensitivity in a principled and computationally feasible way remains challenging in practice. Unfortunately, existing sensitivity methods are rarely applicable in modern Bayesian workflows due to their high computational cost and poor performance in moderate to high dimensions. To address these limitations, we introduce a new approach to global sensitivity analysis based on the Fisher divergence. Our method only requires a set of samples from a reference posterior and the ability to evaluate score functions, making it broadly computationally tractable. Under mild regularity conditions, it controls changes in the whole posterior, and provides a bound on the impact of perturbations on the first two moments. We demonstrate these strengths on challenging Bayesian inference problems which are practically out of reach of existing approaches, including generalised Bayesian inference for unnormalised models, inference in Bayesian models of time series, and neural simulation-based inference.
Related
- Likelihood-Free Schrödinger--Föllmer Sampling for Bayesian Inference with Fixed-Proposal Reuse
- Coherent and robust Bayesian inference under informative sampling via sample-level weight models
- Tractable Inference with Kolmogorov-Arnold Networks
- Direct Message Approximation (DMA): A Consistency-Based Framework for Tractable Approximate Inference on Factor Graphs
- An impartial Bayesian hypothesis test for audit sampling
BibTeX
@misc{odnoblyudova2026computationallytractable,
title = {{A computationally-tractable measure of global sensitivity for sampling-based Bayesian inference}},
author = {Arina Odnoblyudova and Charita Dellaporta and François-Xavier Briol},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2605.28099},
eprint = {2605.28099},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.28099},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A computationally-tractable measure of global sensitivity for sampling-based Bayesian inference
AU - Arina Odnoblyudova
AU - Charita Dellaporta
AU - François-Xavier Briol
PY - 2026
DA - 2026/10/06
PB - arXiv
DO - 10.48550/arxiv.2605.28099
UR - https://arxiv.org/abs/2605.28099
AB - Bayesian inference can often be sensitive to the choice of hyperparameters of the prior or likelihood, yet defining and quantifying this sensitivity in a principled and computationally feasible way remains challenging in practice. Unfortunately, existing sensitivity methods are rarely applicable in modern Bayesian workflows due to their high computational cost and poor performance in moderate to high dimensions. To address these limitations, we introduce a new approach to global sensitivity analysis based on the Fisher divergence. Our method only requires a set of samples from a reference posterior and the ability to evaluate score functions, making it broadly computationally tractable. Under mild regularity conditions, it controls changes in the whole posterior, and provides a bound on the impact of perturbations on the first two moments. We demonstrate these strengths on challenging Bayesian inference problems which are practically out of reach of existing approaches, including generalised Bayesian inference for unnormalised models, inference in Bayesian models of time series, and neural simulation-based inference.
ER -