AI Paper Atlas

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

Isotropic Embedding Perturbations for Robust Vision Language Encoders

Hyesong Choi, Daeun Kim, Song Park, Taekyung Kim, Byeongho Heo, Sangdoo Yun, Dongbo Min, Dongyoon Han

ECCV 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 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
inject perturbation in the embedding space instead of pixel space (ablation)ViT-B fine-tuning · CUBaccuracy: 79.33 -> 80.89Improved
inject perturbation in the embedding space instead of pixel space (ablation)ViT-B fine-tuning · ImageNet-1Ktop-1 accuracy: 80.17 -> 82.25Improved
inject perturbation in the embedding space instead of pixel space (ablation)ViT-B fine-tuning · token perturbation covarianceanisotropy ratio lambda_max/lambda_min: 8.4 -> 1.15Improved
add Aether on top of the standard recipe R_bViT-B · ImageNet-1Ktop-1 accuracy: 81.17 -> 82.25Improved
add AugMix or RandErase on top of R_bViT-B · ImageNet-1Ktop-1 accuracy: 81.17 -> 81.16No clear effect
add Aether on top of R_bCLIP · ImageNet-1Ktop-1 accuracy: 84.10 -> 84.37Improved
add Aether to fine-tuningAIMv2 · ImageNet-1Ktop-1 accuracy: 86.22 -> 87.01Improved
add Aether to fine-tuningSigLIP 2 · ImageNet-1Ktop-1 accuracy: 73.64 -> 73.79No clear effect
add Aether on top of R_bSwinV2-L · ImageNet-1Ktop-1 accuracy: 85.21 -> 85.30No clear effect
add AetherResNet-50 · ImageNet-1Ktop-1 accuracy: 79.86 -> 80.04No clear effect
add Aether on top of R_b in fine-tuningViT-L · ImageNet-1Ktop-1 accuracy: 84.71 -> 85.35Improved
add Aether during fine-tuning of diffusion-based SSL modelDiffMAE · ImageNet-1Ktop-1 accuracy: 82.18 -> 82.50Improved

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