List your company
ResearchFormerly WizWand

Multi-modal Retrieval (Image Query) on MMQA

76.52Recall@20CLIP (ViT-L/14)
-3.060817.599638.2658.9204Nov 27, 2025
Updated 8mo ago

Evaluation results

MethodLinks
CLIP (ViT-L/14)Scoring Method=Ours2025.1176.5258.8863.05
Long-CLIP-LScoring Method=Ours2025.1173.9154.0458.63
Long-CLIP-LScoring Method=Std2025.1171.7449.3854.52
Long-CLIP-BScoring Method=Ours2025.1166.9650.7254.51
CLIP (ViT-B/32)Scoring Method=Ours2025.1166.0945.0349.86
Long-CLIP-BScoring Method=Std2025.1166.0947.6751.94
CLIP (ViT-L/14)Scoring Method=Std2025.1162.1743.3247.7
CLIP (ViT-B/32)Scoring Method=Std2025.1152.6136.6540.32
Cohere Embed 3Scoring Method=Std2025.1152.1727.2432.92
Cohere Embed 3Scoring Method=Ours2025.1152.1728.1733.61
Cohere Embed 3Scoring Method=Cos2025.1150.4320.7927.61
BLIPScoring Method=Ours2025.1143.4827.4531.15
E5-VScoring Method=Std2025.1141.7421.3325.91
E5-VScoring Method=Ours2025.1141.7421.5526.09
BLIPScoring Method=Std2025.1139.5723.9727.54
E5-VScoring Method=Cos2025.1138.717.3422.06
CLIP (ViT-L/14)Scoring Method=Cos2025.111.30.410.62
Long-CLIP-LScoring Method=Cos2025.110.430.110.19
CLIP (ViT-B/32)Scoring Method=Cos2025.11000
Long-CLIP-BScoring Method=Cos2025.11000
BLIPScoring Method=Cos2025.11000