AI Paper Atlas

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

RANKCLIP: Ranking-Consistent Language-Image Pretraining

Yiming Zhang, Zhuokai Zhao, Zhaorun Chen, Zhili Feng, Zenghui Ding, Yining Sun

ICCV 2025 · Vision-language and multimodal › CLIP and zero-shot · approach: CLIP training recipes and data curation · 13 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 list-wise ranking consistency loss to CLIPCLIP ViT-B/32 CC3M · ImageNet1K zero-shotTop-1 accuracy: 9.06% -> 10.16%Improved
add ranking consistency loss with RN50 backboneCLIP RN50 · ImageNet1K zero-shotTop-1 accuracy: 21.6% -> 30.9%Improved
add ranking consistency loss with ViT-B/32 backboneCLIP ViT-B/32 · ImageNet1K zero-shotTop-1 accuracy: 20.7% -> 26.2%Improved
add ranking consistency lossCLIP RN50 · Linear probing averageaccuracy: 64.2% -> 68.9%Improved
add ranking consistency lossCLIP ViT-B/32 · Linear probing averageaccuracy: 60.7% -> 61.3%Improved
add ranking consistency lossCLIP ViT-B/32 CC3M · MSCOCO retrieval Recall@10Recall@10: image-to-text 26.94% -> 28.48%; text-to-image 14.04% -> 14.18%Mixed
add ranking consistency lossCLIP ViT-B/32 CC3M · ImageNet-RTop-1 accuracy: 9.36% -> 11.34%Improved
add ranking consistency lossCLIP ViT-B/32 CC3M · SVHN linear probingaccuracy: 50.5% -> 47.7%Hurt
use cross-modal rank loss only (drop in-modal) (ablation)RankCLIP · ImageNet1K alignment/uniformityuniformity score S_U: -0.13 -> -0.14Mixed
use in-modal rank loss only (drop cross-modal) (ablation)RankCLIP · ImageNet1K alignment/uniformityuniformity score S_U: -0.13 -> -0.17Hurt
add ranking consistency lossCLIP · ImageNet1K alignmentalignment score S_A: 0.33 -> 0.29Hurt
use adaptive lambda schedule instead of fixed weights (ablation)RankCLIP · ImageNet1K and MSCOCOaccuracy and recallImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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