List your company

Compositional Reasoning on CREPE

44.9AccuracyReasonCLIP + READ
13.80421.87729.9538.023Jun 25, 2026

Evaluation results

MethodLinks
ReasonCLIP + READViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0644.9
READ-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0641.5
Long-CLIPViT=L/14, Resolution=224, Data=0.4B2026.0635.1
CE-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0634.8
NegCLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0630.5
ReasonCLIPViT=L/14, Resolution=224, Data=+58M2026.0624.8
ReasonCLIPViT=L/14, Resolution=336, Data=+58M2026.0624.8
ReasonCLIPViT=B/32, Resolution=224, Data=+58M2026.0624
CLIPViT=B/32, Resolution=224, Data=0.4B2026.0623.9
Eva-CLIP-02ViT=L/14, Resolution=224, Data=2.0B2026.0621.4
CLIPViT=L/14, Resolution=336, Data=0.4B2026.0620.9
CLIPViT=L/14, Resolution=224, Data=0.4B2026.0620.5
ReasonSigLIP2ViT=G/16, Resolution=384, Data=+58M2026.0620.5
ReasonSigLIPViT=So/14, Resolution=384, Data=+58M2026.0620.1
MetaCLIPViT=L/14, Resolution=224, Data=2.5B2026.0619.7
MetaCLIPViT=B/32, Resolution=224, Data=0.4B2026.0618.2
SigLIPViT=So/14, Resolution=384, Data=10.0B2026.0618
GNM-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0617.4
SigLIP2ViT=G/16, Resolution=384, Data=10.0B2026.0616.5
SigLIP2ViT=L/16, Resolution=384, Data=10.0B2026.0615.9
Triplet-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.0615