A Trust Crisis In Simulation-Based Inference? Your Posterior Approximations Can Be Unfaithful
Not on the map. Why?
Abstract (arXiv)
We present extensive empirical evidence showing that current Bayesian simulation-based inference algorithms can produce computationally unfaithful posterior approximations. Our results show that all benchmarked algorithms -- (Sequential) Neural Posterior Estimation, (Sequential) Neural Ratio Estimation, Sequential Neural Likelihood and variants of Approximate Bayesian Computation -- can yield overconfident posterior approximations, which makes them unreliable for scientific use cases and falsificationist inquiry. Failing to address this issue may reduce the range of applicability of simulation-based inference. For this reason, we argue that research efforts should be made towards theoretical and methodological developments of conservative approximate inference algorithms and present research directions towards this objective. In this regard, we show empirical evidence that ensembling posterior surrogates provides more reliable approximations and mitigates the issue.
Key facts. Cited by 28 papers in the map (data through 14 Oct 2026). Builds on 28 of its 74 references that are in the map. First posted on arXiv 13 Oct 2021.
Builds on
28 of 74 references.
References · 28
- 2014Adam: A Method for Stochastic Optimizationcited by 37,240
- 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,926
- 2015Variational Inference with Normalizing Flowscited by 2,170
- 2016Density estimation using Real NVPcited by 2,042
- 2014NICE: Non-linear Independent Components Estimationcited by 1,058
- 2017Self-Normalizing Neural Networkscited by 598
- 2019The frontier of simulation-based inferencecited by 502
- 2016Revisiting Classifier Two-Sample Testscited by 268
- 2018Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flowscited by 221
- 2019Automatic Posterior Transformation for Likelihood-Free Inferencecited by 219
- 2017Flexible statistical inference for mechanistic models of neural dynamicscited by 163
- 2021Benchmarking Simulation-Based Inferencecited by 162
16 more
- 2015Training generative neural networks via Maximum Mean Discrepancy optimizationcited by 151
- 2018Validating Bayesian Inference Algorithms with Simulation-Based Calibrationcited by 124
- 2020On Contrastive Learning for Likelihood-free Inferencecited by 98
- 2015Approximating Likelihood Ratios with Calibrated Discriminative Classifierscited by 90
- 2016Likelihood-free inference by ratio estimationcited by 81
- 2021Robust Generalised Bayesian Inference for Intractable Likelihoodscited by 53
- 2021Diagnostics for Conditional Density Models and Bayesian Inference Algorithmscited by 28
- 2018A Guide to Constraining Effective Field Theories with Machine Learningcited by 25
- 2020Score Matched Neural Exponential Families for Likelihood-Free Inferencecited by 21
- 2020Generalized Posteriors in Approximate Bayesian Computationcited by 19
- 2020Confidence Sets and Hypothesis Testing in a Likelihood-Free Inference Settingcited by 18
- 2019Mining for Dark Matter Substructure: Inferring subhalo population properties from strong lenses with machine learningcited by 15
- 2020Towards constraining warm dark matter with stellar streams through neural simulation-based inferencecited by 14
- 2013Diagnostic tools of approximate Bayesian computation using the coverage propertycited by 10
- 2021Arbitrary Marginal Neural Ratio Estimation for Simulation-based Inferencecited by 5
- 2013Bounding the Test Log-Likelihood of Generative Modelscited by 2
Built on by
Citing papers per year: 2022: 1; 2023: 6; 2024: 7; 2025: 5; 2026: 9.
28 papers.
Sort:
By area:
- 2024All-in-one simulation-based inferencecited by 74
- 2022Sequential Neural Score Estimation: Likelihood-Free Inference with Conditional Score Based Diffusion Modelscited by 50
- 2023JANA: Jointly Amortized Neural Approximation of Complex Bayesian Modelscited by 41
- 2023Adversarial robustness of amortized Bayesian inferencecited by 17
- 2023Hierarchical Neural Simulation-Based Inference Over Event Ensemblescited by 9
- 2025G-Sim: Generative Simulations with Large Language Models and Gradient-Free Calibrationcited by 8
- 2023AI for Science: An Emerging Agendacited by 5
- 2024Simulation-Based Inference with Quantile Regressioncited by 4
- 2024Conformal Calibration of Statistical Confidence Setscited by 3
- 2023Discriminative calibration: Check Bayesian computation from simulations and flexible classifiercited by 2
- 2026From Mice to Trains: Amortized Bayesian Inference on Graph Datacited by 2
- 2024SoftCVI: Contrastive variational inference with self-generated soft labelscited by 1
16 more
- 2024Physically Interpretable Probabilistic Domain Characterizationcited by 1
- 2025A Review of Diffusion-based Simulation-Based Inference: Foundations and Applications in Non-Ideal Data Scenarioscited by 1
- 2026Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problemscited by 1
- 2023An efficient likelihood-free Bayesian inference method based on sequential neural posterior estimationcited by 0
- 2024Leveraging Nested MLMC for Sequential Neural Posterior Estimation with Intractable Likelihoodscited by 0
- 2024Cost Estimation in Unit Commitment Problems Using Simulation-Based Inferencecited by 0
- 2025Fast reconstruction of degenerate populations of conductance-based neuron models from spike timescited by 0
- 2025Dark Energy Survey Year 3 results: Simulation-based wCDM inference from weak lensing and galaxy clustering maps with deep learning: Analysis designcited by 0
- 2025Fast and Robust Simulation-Based Inference With Optimization Monte Carlocited by 0
- 2026Simulation-Based Inference via Regression Projection and Batched Discrepanciescited by 0
- 2026Confidence Freeze: Early Success Induces a Metastable Decoupling of Metacognition and Behaviourcited by 0
- 2026Conservative neural posterior estimation via distributionally robust trainingcited by 0
- 2026Amortized Inference for Sampling Distributions Where the Bootstrap Failscited by 0
- 2026Priors learned from legacy reconstructions inherit undetectable overconfidencecited by 0
- 2026Bayesian Uncertainty Quantification for fMRI Functional Connectivity via Simulation-Based Inferencecited by 0
- 2026A Missing Latent, Not a Missing Simulator: Radius-Augmented Inference for Real JWST Retrievalcited by 0
- 2026A Missing Latent, Not a Missing Simulator: Radius-Augmented Inference for Real JWST Retrieval30 Sep 2026
- 2026Bayesian Uncertainty Quantification for fMRI Functional Connectivity via Simulation-Based Inference24 Sep 2026
- 2026Priors learned from legacy reconstructions inherit undetectable overconfidence23 Jul 2026
- 2026Amortized Inference for Sampling Distributions Where the Bootstrap Fails18 Jul 2026
- 2026Conservative neural posterior estimation via distributionally robust training27 May 2026
- 2026Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problems21 May 2026
- 2026Confidence Freeze: Early Success Induces a Metastable Decoupling of Metacognition and Behaviour22 Mar 2026
- 2026Simulation-Based Inference via Regression Projection and Batched Discrepancies3 Feb 2026
- 2026From Mice to Trains: Amortized Bayesian Inference on Graph Data5 Jan 2026
- 2025A Review of Diffusion-based Simulation-Based Inference: Foundations and Applications in Non-Ideal Data Scenarios26 Dec 2025
- 2025Fast and Robust Simulation-Based Inference With Optimization Monte Carlo17 Nov 2025
- 2025Dark Energy Survey Year 3 results: Simulation-based wCDM inference from weak lensing and galaxy clustering maps with deep learning: Analysis design6 Nov 2025
16 more
- 2025Fast reconstruction of degenerate populations of conductance-based neuron models from spike times16 Sep 2025
- 2025G-Sim: Generative Simulations with Large Language Models and Gradient-Free Calibration10 Jun 2025
- 2024Conformal Calibration of Statistical Confidence Sets28 Nov 2024
- 2024Physically Interpretable Probabilistic Domain Characterization22 Nov 2024
- 2024Cost Estimation in Unit Commitment Problems Using Simulation-Based Inference5 Sep 2024
- 2024SoftCVI: Contrastive variational inference with self-generated soft labels22 Jul 2024
- 2024All-in-one simulation-based inference15 Apr 2024
- 2024Leveraging Nested MLMC for Sequential Neural Posterior Estimation with Intractable Likelihoods30 Jan 2024
- 2024Simulation-Based Inference with Quantile Regression4 Jan 2024
- 2023An efficient likelihood-free Bayesian inference method based on sequential neural posterior estimation21 Nov 2023
- 2023Hierarchical Neural Simulation-Based Inference Over Event Ensembles21 Jun 2023
- 2023Adversarial robustness of amortized Bayesian inference24 May 2023
- 2023Discriminative calibration: Check Bayesian computation from simulations and flexible classifier24 May 2023
- 2023AI for Science: An Emerging Agenda7 Mar 2023
- 2023JANA: Jointly Amortized Neural Approximation of Complex Bayesian Models17 Feb 2023
- 2022Sequential Neural Score Estimation: Likelihood-Free Inference with Conditional Score Based Diffusion Models10 Oct 2022
- 2024All-in-one simulation-based inferencecited by 74
- 2023JANA: Jointly Amortized Neural Approximation of Complex Bayesian Modelscited by 41
- 2023Adversarial robustness of amortized Bayesian inferencecited by 17
- 2023Hierarchical Neural Simulation-Based Inference Over Event Ensemblescited by 9
- 2025G-Sim: Generative Simulations with Large Language Models and Gradient-Free Calibrationcited by 8
- 2023AI for Science: An Emerging Agendacited by 5
- 2024Simulation-Based Inference with Quantile Regressioncited by 4
- 2024Conformal Calibration of Statistical Confidence Setscited by 3
- 2023Discriminative calibration: Check Bayesian computation from simulations and flexible classifiercited by 2
- 2026From Mice to Trains: Amortized Bayesian Inference on Graph Datacited by 2
- 2024SoftCVI: Contrastive variational inference with self-generated soft labelscited by 1
- 2024Physically Interpretable Probabilistic Domain Characterizationcited by 1
12 most cited of 26 All 26 (JSON)
- 2026A Missing Latent, Not a Missing Simulator: Radius-Augmented Inference for Real JWST Retrieval30 Sep 2026
- 2026Bayesian Uncertainty Quantification for fMRI Functional Connectivity via Simulation-Based Inference24 Sep 2026
- 2026Priors learned from legacy reconstructions inherit undetectable overconfidence23 Jul 2026
- 2026Amortized Inference for Sampling Distributions Where the Bootstrap Fails18 Jul 2026
- 2026Conservative neural posterior estimation via distributionally robust training27 May 2026
- 2026Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problems21 May 2026
- 2026Confidence Freeze: Early Success Induces a Metastable Decoupling of Metacognition and Behaviour22 Mar 2026
- 2026Simulation-Based Inference via Regression Projection and Batched Discrepancies3 Feb 2026
- 2026From Mice to Trains: Amortized Bayesian Inference on Graph Data5 Jan 2026
- 2025Fast and Robust Simulation-Based Inference With Optimization Monte Carlo17 Nov 2025
- 2025Dark Energy Survey Year 3 results: Simulation-based wCDM inference from weak lensing and galaxy clustering maps with deep learning: Analysis design6 Nov 2025
- 2025Fast reconstruction of degenerate populations of conductance-based neuron models from spike times16 Sep 2025
12 newest of 26 All 26 (JSON)