Saved in:
Bibliographic Details
Main Authors: Gao, Zijun, Sun, Yan, Su, Han
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.18897
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914108373204992
author Gao, Zijun
Sun, Yan
Su, Han
author_facet Gao, Zijun
Sun, Yan
Su, Han
contents Generative models have achieved remarkable success across a range of applications, yet their evaluation still lacks principled uncertainty quantification. In this paper, we develop a method for comparing how close different generative models are to the underlying distribution of test samples. Particularly, our approach employs the Kullback-Leibler (KL) divergence to measure the distance between a generative model and the unknown test distribution, as KL requires no tuning parameters such as the kernels used by RKHS-based distances, and is the only $f$-divergence that admits a crucial cancellation to enable the uncertainty quantification. Furthermore, we extend our method to comparing conditional generative models and leverage Edgeworth expansions to address limited-data settings. On simulated datasets with known ground truth, we show that our approach realizes effective coverage rates, and has higher power compared to kernel-based methods. When applied to generative models on image and text datasets, our procedure yields conclusions consistent with benchmark metrics but with statistical confidence.
format Preprint
id arxiv_https___arxiv_org_abs_2501_18897
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Statistical Inference for Generative Model Comparison
Gao, Zijun
Sun, Yan
Su, Han
Machine Learning
Generative models have achieved remarkable success across a range of applications, yet their evaluation still lacks principled uncertainty quantification. In this paper, we develop a method for comparing how close different generative models are to the underlying distribution of test samples. Particularly, our approach employs the Kullback-Leibler (KL) divergence to measure the distance between a generative model and the unknown test distribution, as KL requires no tuning parameters such as the kernels used by RKHS-based distances, and is the only $f$-divergence that admits a crucial cancellation to enable the uncertainty quantification. Furthermore, we extend our method to comparing conditional generative models and leverage Edgeworth expansions to address limited-data settings. On simulated datasets with known ground truth, we show that our approach realizes effective coverage rates, and has higher power compared to kernel-based methods. When applied to generative models on image and text datasets, our procedure yields conclusions consistent with benchmark metrics but with statistical confidence.
title Statistical Inference for Generative Model Comparison
topic Machine Learning
url https://arxiv.org/abs/2501.18897