![[2602.16449] GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16449] GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation
Summary
The paper presents GICDM, a method to mitigate hubness in distance-based evaluations of generative models, enhancing reliability and alignment with human judgment.
Why It Matters
Understanding and addressing hubness is critical in machine learning, particularly for generative models, as it directly impacts the accuracy of distance-based evaluations. GICDM offers a solution that improves model evaluation metrics, which is essential for advancing AI applications.
Key Takeaways
- Hubness can distort nearest neighbor relationships in high-dimensional spaces.
- GICDM corrects neighborhood estimation for both real and generated data.
- The method includes a multi-scale extension to enhance empirical performance.
- Experiments show GICDM resolves hubness-induced failures.
- Improved evaluation metrics align better with human judgment.
Computer Science > Machine Learning arXiv:2602.16449 (cs) [Submitted on 18 Feb 2026] Title:GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation Authors:Nicolas Salvy, Hugues Talbot, Bertrand Thirion View a PDF of the paper titled GICDM: Mitigating Hubness for Reliable Distance-Based Generative Model Evaluation, by Nicolas Salvy and 1 other authors View PDF HTML (experimental) 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 judgment. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) Cite as: arXiv:2602.16449 [cs.LG] (or arXiv:2602.16449v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.16449 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Nicolas Sal...
