MIRA: A Score for Conditional Distribution Accuracy and Model Comparison
Not on the map. Why?
Abstract (arXiv)
We introduce Mira, a sample-based score for assessing the accuracy of a candidate conditional distribution using only joint samples from the true data-generating process. Relying on the principle that distributions coincide if they assign equal probability mass to all regions, we derive an analytic expression for the Mira statistic, whose average defines the Mira score. This formulation further allows us to compute theoretical reference values and uncertainty estimates when the candidate distribution matches the true one. This framework enables model comparison by quantifying the alignment between the conditional distribution of a candidate model and the true data generating process. Consequently, Mira enables Bayesian model comparison through direct posterior validation, bypassing the challenging evidence computation. We demonstrate its effectiveness across several toy problems and Bayesian inference tasks.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 41 of its 84 references that are in the map. First posted on arXiv 3 May 2026.
Builds on
41 of 84 references.
References · 41
- 2020Denoising Diffusion Probabilistic Modelscited by 20,851
- 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
- 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
- 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,375
- 2016f-GAN: Training Generative Neural Samplers using Variational Divergence Minimizationcited by 504
- 2018Do Deep Generative Models Know What They Don't Know?cited by 443
- 2018Assessing Generative Models via Precision and Recallcited by 401
- 2023Denoising Diffusion Models for Plug-and-Play Image Restorationcited by 335
- 2017Deep Models Under the GAN: Information Leakage from Collaborative Deep Learningcited by 326
- 2022MCVD: Masked Conditional Video Diffusion for Prediction, Generation, and Interpolationcited by 307
- 2015A note on the evaluation of generative modelscited by 291
- 2022Brain Imaging Generation with Latent Diffusion Modelscited by 267
29 more
- 2016Revisiting Classifier Two-Sample Testscited by 258
- 2023A Variational Perspective on Solving Inverse Problems with Diffusion Modelscited by 217
- 2023Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion modelscited by 199
- 2021How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Modelscited by 193
- 2024Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealingcited by 157
- 2023Score-Based Diffusion Models as Principled Priors for Inverse Imagingcited by 126
- 2018Validating Bayesian Inference Algorithms with Simulation-Based Calibrationcited by 118
- 2024Principled Probabilistic Imaging using Diffusion Models as Plug-and-Play Priorscited by 93
- 2019X2CT-GAN: Reconstructing CT from Biplanar X-Rays with Generative Adversarial Networkscited by 85
- 2018Variational Autoencoder with Arbitrary Conditioningcited by 68
- 2023Sampling-Based Accuracy Testing of Posterior Estimators for General Inferencecited by 45
- 2020Perfect density models cannot guarantee anomaly detectioncited by 39
- 2019How to GAN away Detector Effectscited by 28
- 2020Amortized Bayesian model comparison with evidential deep learningcited by 22
- 2022Posterior samples of source galaxies in strong gravitational lenses with score-based priorscited by 20
- 2022Probabilistic Mass Mapping with Neural Score Estimationcited by 17
- 2022Simulator-Based Inference with Waldo: Confidence Regions by Leveraging Prediction Algorithms and Posterior Estimators for Inverse Problemscited by 14
- 2023Improving Generative Model-based Unfolding with Schr"odinger Bridgescited by 13
- 2023Score-based Data Assimilation for a Two-Layer Quasi-Geostrophic Modelcited by 11
- 2020Empirical Analysis of Overfitting and Mode Drop in GAN Trainingcited by 10
- 2023Evidence Networks: simple losses for fast, amortized, neural Bayesian model comparisoncited by 10
- 2022Towards Instance-Wise Calibration: Local Amortized Diagnostics and Reshaping of Conditional Densities (LADaR)cited by 9
- 2024PQMass: Probabilistic Assessment of the Quality of Generative Models using Probability Mass Estimationcited by 8
- 2024When Diffusion MRI Meets Diffusion Model: A Novel Deep Generative Model for Diffusion MRI Generationcited by 6
- 2023Discovering Structure From Corruption for Unsupervised Image Reconstructioncited by 5
- 2023Variational Inference with Coverage Guarantees in Simulation-Based Inferencecited by 5
- 2024Tackling the Problem of Distributional Shifts: Correcting Misspecified, High-Dimensional Data-Driven Priors for Inverse Problemscited by 5
- 2025Training-Free Data Assimilation with GenCastcited by 5
- 2025STeP: A Framework for Solving Scientific Video Inverse Problems with Spatiotemporal Diffusion Priorscited by 3
Built on by
Citing papers per year: 2026: 1.
1 paper.