AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Emergent Visual-Semantic Hierarchies in Image-Text Representations

Morris Alper, Hadar Averbuch-Elor

ECCV 2024 · Vision-language and multimodal › CLIP and zero-shot · approach: Recaptioning and text-supervised pretraining · 12 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
fine-tune with RE contrastive loss on text-only HierarCapsCLIP ViT-B · HierarCaps testrecall: 0.36 -> 0.47Improved
fine-tune with RE lossCLIP ViT-B · HierarCaps testtau_d: 0.89 -> 0.99Improved
fine-tune with RE lossOpenCLIP ViT-H · HyperLexSpearman rho_all: 0.06 -> 0.37Improved
fine-tune with RE loss on hierarchy dataCLIP ViT-B · MS-COCO text-to-imageR@1: 0.30 -> 0.31No clear effect
fine-tune with RE lossCLIP ViT-B · CIFAR-100 zero-shotaccuracy: 0.66 -> 0.65No clear effect
use radial embedding contrastive loss instead of entailment cone loss (ablation)CLIP ViT-B fine-tuning · HierarCaps testrecall: 0.30 -> 0.47Improved
use pretrained foundation VLM instead of from-scratch hyperbolic MERUCLIP ViT-L · HierarCaps testtau_d: MERU 0.79 -> CLIP-L FT 0.97Improved
use text-only RE fine-tuning compared to dedicated lexical entailment modelsCLIP ViT-B · HyperLexSpearman rho_all: 0.51 vs LEAR 0.69Hurt

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