AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Cross-Modal Redundancy and the Geometry of Vision–Language Embeddings

Grégoire DHIMOÏLA, Thomas Fel, Victor Boutin, Agustin Martin Picard

ICLR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 3 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 Iso-Energy alignment loss (ablation)Matching Pursuit SAE, CLIP · LAIONfunctional alignment rho: 0.327 -> 4.232Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE, OpenCLIP · LAIONfunctional alignment rho: 4.072 -> 16.02Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE, SigLIP · LAIONFDA: 8.831 -> 34.95Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE, CLIP · LAIONprobing accuracy: 0.847 -> 0.915Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE, CLIP · LAIONinterventional robustness delta_r: 0.224 -> 0.125Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE, CLIP · LAIONreconstruction R2: 0.859 -> 0.837No clear effect
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE · synthetic, Iso-Energy holdsWasserstein distance: 0.396 -> 0.184Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE · synthetic, Iso-Energy holdsmean matching accuracy: 0.29 -> 0.52Improved
add Iso-Energy alignment loss (ablation)Matching Pursuit SAE · synthetic, Iso-Energy violatedWasserstein distance: 0.19No clear effect
restrict vector arithmetic to bimodal atomsCLIP query · FashionIQOOD score: 0.97 -> 0.77Improved
restrict vector arithmetic to bimodal atomsSigLIP2 query · FashionIQOOD score: 0.99 -> 0.61Improved
filter unimodal atoms with binary maskCLIP embeddings · modality gapmodality gapImproved

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