GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation

Nicolas Salvy, Hugues Talbot, Bertrand Thirion
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:107378-107411, 2026.

Abstract

Generative model evaluation commonly relies on high-dimensional embedding spaces to compute distances between samples. We show that dataset representations in these spaces are affected by the hubness phenomenon, which distorts nearest-neighbor relationships and biases distance-based metrics. Building on the classical Iterative Contextual Dissimilarity Measure (ICDM), we introduce Generative ICDM (GICDM), a method to correct neighborhood estimation for both real and generated data. We introduce a multi-scale extension to improve empirical behavior. Extensive experiments on synthetic and real benchmarks demonstrate that GICDM resolves hubness-induced failures, restores reliable metric behavior, and improves alignment with human assessment.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-salvy26a, title = {{GICDM}: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation}, author = {Salvy, Nicolas and Talbot, Hugues and Thirion, Bertrand}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {107378--107411}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/salvy26a/salvy26a.pdf}, url = {https://proceedings.mlr.press/v306/salvy26a.html}, abstract = {Generative model evaluation commonly relies on high-dimensional embedding spaces to compute distances between samples. We show that dataset representations in these spaces are affected by the hubness phenomenon, which distorts nearest-neighbor relationships and biases distance-based metrics. Building on the classical Iterative Contextual Dissimilarity Measure (ICDM), we introduce Generative ICDM (GICDM), a method to correct neighborhood estimation for both real and generated data. We introduce a multi-scale extension to improve empirical behavior. Extensive experiments on synthetic and real benchmarks demonstrate that GICDM resolves hubness-induced failures, restores reliable metric behavior, and improves alignment with human assessment.} }
Endnote
%0 Conference Paper %T GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation %A Nicolas Salvy %A Hugues Talbot %A Bertrand Thirion %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-salvy26a %I PMLR %P 107378--107411 %U https://proceedings.mlr.press/v306/salvy26a.html %V 306 %X Generative model evaluation commonly relies on high-dimensional embedding spaces to compute distances between samples. We show that dataset representations in these spaces are affected by the hubness phenomenon, which distorts nearest-neighbor relationships and biases distance-based metrics. Building on the classical Iterative Contextual Dissimilarity Measure (ICDM), we introduce Generative ICDM (GICDM), a method to correct neighborhood estimation for both real and generated data. We introduce a multi-scale extension to improve empirical behavior. Extensive experiments on synthetic and real benchmarks demonstrate that GICDM resolves hubness-induced failures, restores reliable metric behavior, and improves alignment with human assessment.
APA
Salvy, N., Talbot, H. & Thirion, B.. (2026). GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:107378-107411 Available from https://proceedings.mlr.press/v306/salvy26a.html.

Related Material