AI Paper Atlas

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

Explaining CLIP Zero-shot Predictions Through Concepts

Onat Ozdemir, Anders Christensen, Stephan Alaniz, Zeynep Akata, Emre Akbas

CVPR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Attribute-based zero-shot learning · 1 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
project CLIP into a learned concept space (EZPC) instead of plain CLIPCLIP RN50 · ImageNet-100 GZSharmonic mean H: 0.693 -> 0.682No clear effect
project CLIP into a learned concept space instead of plain CLIPCLIP RN50 · ImageNet-1k GZSharmonic mean H: 0.530 -> 0.481Hurt
use EZPC instead of Z-CBM and SpLiCEconcept-based zero-shot CLIP · CUB GZSharmonic mean H: 0.189 -> 0.465 vs Z-CBMImproved
use EZPC instead of Z-CBMconcept-based zero-shot CLIP · Places365 GZSharmonic mean H: 0.357 -> 0.352No clear effect
increase reconstruction weight lambda from 0.01 to 10 (ablation)EZPC · ImageNet-100 GZSharmonic mean H: 0.358 -> 0.695Improved
increase lambda beyond 10 (to 100 or 1000) (ablation)EZPC · ImageNet-100 GZSharmonic mean H: 0.695 -> 0.694No clear effect
increase lambda to 100 for explanations (ablation)EZPC · qualitative concept relevanceexplanation qualityHurt
use a larger backbone (ViT-L/14, SigLIP) for EZPC (ablation)EZPC · ImageNet-100 GZSharmonic mean H: 0.682 (RN50) -> 0.878 (SigLIP)Improved
train EZPC on ImageNet-100 and test on CIFAR-100EZPC vs CLIP · CIFAR-100 cross-dataset GZSharmonic mean H: 0.380 -> 0.409Improved
train EZPC on ImageNet-100 and test on CUBEZPC vs CLIP · CUB cross-dataset GZSharmonic mean H: 0.526 -> 0.515No clear effect
replace per-image optimization (SpLiCE, Z-CBM) with single linear projectionEZPC · ImageNet-100 inferencelatency (ms/img): 338.51 / 542.34 -> 5.90Improved

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