List your company

Image-Text Retrieval on MSCOCO (R@1, R@5, nDCG@10)

53.4Recall@1DFN-CLIP
18.87227.83636.845.764Jul 6, 2026

Evaluation results

MethodLinks
DFN-CLIPArchitecture=Single-vector2026.0753.477.869
SigLIP2Architecture=Single-vector2026.0752.175.266.9
SaMer (w/ FT)Architecture=Multi-vector, Backbone=ColPali, Token Compression (K)=64, Compression Strategy=Merging, Adapted=true2026.0751.675.567.1
EVA-CLIPArchitecture=Single-vector2026.075174.666.1
OpenCLIPArchitecture=Single-vector2026.0748.573.264.4
SaMer (w/ FT)Architecture=Multi-vector, Backbone=ColQwen2, Token Compression (K)=64, Compression Strategy=Merging, Adapted=true2026.0747.572.363.5
ColPaliArchitecture=Multi-vector, Backbone=ColPali, Compression=None2026.0747.471.763.3
GMEArchitecture=VLM-based2026.0747.371.162.7
SaMer (w/o FT)Architecture=Multi-vector, Backbone=ColPali, Token Compression (K)=64, Compression Strategy=Merging, Adapted=false2026.0745.169.761.1
H-PoolArchitecture=Multi-vector, Backbone=ColPali, Token Compression (K)=64, Compression Strategy=Pooling2026.0744.969.660.9
ColQwen2Architecture=Multi-vector, Backbone=ColQwen2, Compression=None2026.0743.36859.5
VisRAGArchitecture=VLM-based2026.0742.567.558.9
H-PoolArchitecture=Multi-vector, Backbone=ColQwen2, Token Compression (K)=64, Compression Strategy=Pooling2026.0742.267.258.6
SaMer (w/o FT)Architecture=Multi-vector, Backbone=ColQwen2, Token Compression (K)=64, Compression Strategy=Merging, Adapted=false2026.0742.16758.4
SAPArchitecture=Multi-vector, Backbone=ColQwen2, Token Compression (K)=64, Compression Strategy=Structural anchor pruning2026.0740.765.857.2
HPCArchitecture=Multi-vector, Backbone=ColQwen2, Token Compression (K)=64, Compression Strategy=Attention-based pruning2026.0740.164.856.5
SAPArchitecture=Multi-vector, Backbone=ColPali, Token Compression (K)=64, Compression Strategy=Structural anchor pruning2026.0739.663.755.4
MetaCLIPArchitecture=Single-vector2026.0735.761.752.8
CLIPArchitecture=Single-vector2026.0730.85647.7
HPCArchitecture=Multi-vector, Backbone=ColPali, Token Compression (K)=64, Compression Strategy=Attention-based pruning2026.0730.652.545.4
VLM2Vec-v2Architecture=VLM-based2026.0720.244.136.5