List your company

Long-caption Cross-modal Retrieval on DOCCI (test)

81.12T2I Recall@1HyFL-CLIP
27.081641.110855.1469.1692Jul 1, 2026

Evaluation results

MethodLinks
HyFL-CLIPModel Type=Euclidean VLM with our hyperbolic fine-tuning, Backbone=ViT-B, Protocol=Zero-shot2026.0781.1278.41
UNCHA + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0763.5766.92
HyCoCLIP + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0761.1663.73
HyFL-CLIPModel Type=Euclidean VLM with our hyperbolic fine-tuning, Backbone=ViT-B, Protocol=Zero-shot2026.0758.7559
MERU + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0758.161.71
OpenCLIPModel Type=Euclidean VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0757.2260.84
HyCoCLIPModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0752.0453.78
MERUModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0749.853.16
UNCHAModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0746.7646.73
UNCHA + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0744.2347.03
HyCoCLIP + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0742.7445.3
MERU + OursModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot, Framework=hyperbolic fine-tuning without distillation loss2026.0741.0844.75
OpenCLIPModel Type=Euclidean VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0733.7336.83
HyCoCLIPModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0732.633.39
MERUModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0732.1134
UNCHAModel Type=State-of-the-art hyperbolic VLM, Backbone=ViT-B, Protocol=Zero-shot2026.0729.1629.27