AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

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

Nicolas Salvy, Hugues Talbot, Bertrand Thirion

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Equivariant and disentangled latent representations · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add GICDM hubness correctionClipped Density · Synthetic benchmark (Purpose)passed tests out of 14: 8/14 -> 10/14Improved
add GICDM hubness correctionClipped Density · Synthetic benchmark (Bounds)passed tests out of 13: 8/13 -> 11/13Improved
add GICDM hubness correctionClipped Coverage · Synthetic benchmark (Purpose)passed tests out of 14: 8/14 -> 10/14Improved
add GICDM hubness correctionClipped Coverage · Synthetic benchmark (Bounds)passed tests out of 13: 9/13 -> 11/13Improved
add GICDM hubness correctionClipped Density · Synthetic benchmark (all criteria)passed criteria: 18 -> 23Improved
add GICDM hubness correctionClipped Coverage · Synthetic benchmark (all criteria)passed criteria: 18 -> 22Improved
add GICDM hubness correctionClipped Density (DINOv3) · CIFAR-10 human correlationPearson correlation: 0.82 -> 0.95Improved
add GICDM hubness correctionClipped Density (DINOv3) · ImageNet human correlationPearson correlation: 0.67 -> 0.82Improved
add GICDM hubness correctionClipped Density (DINOv2) · CIFAR-10 human correlationPearson correlation: 0.93 -> 0.97Improved
add GICDM hubness correctionClipped Density (DINOv2) · LSUN Bedroom human correlationPearson correlation: 0.81 -> 0.80No clear effect
add GICDM hubness correctionClipped Coverage (DiT-XL-2, DINOv2) · ImageNet classifier-free guidanceClipped Coverage with guidance: 0.64 -> 0.83 without GICDM; 0.79 -> 0.55 with GICDMImproved
replace NICDM with iterative ICDM (ablation)hubness reduction · 5 datasets/embeddings avgh_1^5(1%): 3.1 -> 1.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy