SOURCE-LINKED INTELLIGENCE

Isotropic Embedding Perturbations for Robust Vision Language Encoders

arXiv · AI, language, vision and robotics · article · Sep 9, 2026 · UTC

Data augmentation is fundamental to training modern deep vision and multimodal models. While individual methods, such as RandAug, CutMix, Mixup, RandErase, and DropPath, offer strong regularization effects, their combined use has saturated in performance due to overlapping functionalities, and aggressive pixel-level manipulations may disrupt delicate cross-modal alignment. This saturation motivates the search for a new augmentation axis within the embedding space rather than the input space. We introduce Aether, a simple plug-in method that applies diffusion-style random perturbations in the e

Read original source ↗ Open in workspace

recordType

paper

region

Global