AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Global Geometry Is Not Enough for Vision Representations

Jiwan Chung, Seon Joo Kim

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 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
measure Jacobian effective rank instead of geometric metrics (G.Iso)encoder diagnostics · Attribute BindingPearson r with binding: +0.11 -> +0.69Improved
compute local isotropy with k=8 instead of k=32 (ablation)local isotropy · Attribute BindingPearson r with binding: +0.045 vs +0.017No clear effect
use effective rank instead of variance formulation for local geometry (ablation)local isotropy · Attribute BindingPearson r with binding: +0.020 -> -0.031No clear effect
evaluate JER on Gaussian noise instead of natural images (ablation)JER · Attribute BindingPearson r with binding: +0.69 -> +0.62Hurt
evaluate JER on random-init weights (ablation)JER · Attribute BindingPearson r with binding: +0.69 -> -0.16Hurt
add Same/Different accuracy to JER in regressionJER · Attribute BindingR^2: 0.47 -> 0.78Improved
use kNN or local PCA readout instead of cosine similarity (ablation)frozen embeddings · Attribute Bindingbinding accuracyHurt
control for log param count in JER-binding correlation (ablation)JER · Attribute BindingPearson r: +0.69 -> +0.54No clear effect
control for ImageNet accuracy and architecture in JER-SugarCrepe (ablation)JER · SugarCrepePearson r: +0.69 -> +0.70No clear effect
control for JER and architecture in ImageNet-SugarCrepe (ablation)ImageNet accuracy · SugarCrepePearson r: +0.67 -> +0.34Hurt
compare pairwise CKA and Procrustes distances with binding differences (ablation)inter-model shape metrics · Attribute BindingMantel r: +0.014 and +0.144No clear effect

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