AI Paper Atlas

Research map / Benchmarks / ImageNet zero-shot

ImageNet zero-shot leaderboard

36 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on ImageNet zero-shot. Best accuracy: 88.31 (SUE, NeurIPS 2025).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

ImageNet zero-shot · accuracy

36 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
188.31SUELearning Shared Representations from Unpaired Data
NeurIPS 2025 · Zero-shot results table
Cross-modal learning
284.1LIMoE-H/14Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts
NeurIPS 2022 · Scaling comparison table
CLIP and zero-shot
383.5Llip64
ViT-G/14
Modeling Caption Diversity in Contrastive Vision-Language Pretraining
ICML 2024 · Zero-shot classification bench
CLIP and zero-shot
482.9ViTamin-XLViTamin: Designing Scalable Vision Models in the Vision-Language Era
CVPR 2024 · Table 3
CLIP and zero-shot
582.7CoBIT-LargeCoBIT: A Contrastive Bi-directional Image-Text Generation Model
ICLR 2024 · Zero-shot Evaluation table
CLIP and zero-shot
682.1MetaCLIP (2.5B)
ViT-bigG/14
Demystifying CLIP Data
ICLR 2024 · Table 2
CLIP and zero-shot
781uCAP (10)
CLIPA ViT-H/14
uCAP: An Unsupervised Prompting Method for Vision-Language Models
ECCV 2024 · Table 1
CLIP and zero-shot
879.64Ours (hierarchical comparisons)
CLIP ViT-L/14
ChatGPT-Powered Hierarchical Comparisons for Image Classification
NeurIPS 2023 · Table 1
CLIP and zero-shot
979.2DataComp-1B
ViT-L/14
DataComp: In search of the next generation of multimodal datasets
NeurIPS 2023 · Table 1
CLIP and zero-shot
1077.97OpenCLIP ViT-H/14, LAION-2B, 34B samples
ViT-H/14
Reproducible Scaling Laws for Contrastive Language-Image Learning
CVPR 2023 · Appendix, ImageNet detailed re
CLIP and zero-shot
26 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.