List your company

Language Compositionality on SugarCrepe (test)

100Replace: Object (R@1)Human
83.46487.75792.0596.343Dec 5, 2024Dec 22, 2024Jan 9, 2025Jan 27, 2025Feb 13, 2025Mar 3, 2025Mar 21, 2025

Evaluation results

MethodLinks
HumanParams (B)=-2024.121009997991009999———
VladVA (Ours)Params (B)=7.06, Backbone=LLaVA-1.5-7B2024.1298.192.186.87982.995.295.8———
BLIP2 (ViT-L)Params (B)=1.17, Backbone=ViT-L2024.1297.681.777.862.165.592.487.4———
EVA-CLIP (18B)Params (B)=18.32024.1297.588.876.165.37692.485———
VLM2Vec (Mistral-7B)Params (B)=7.30, Backbone=Mistral-7B2024.1297.28981.762.972.594.788.6———
EVA-02-CLIP (ViT-E/14+)Params (B)=5.04, Backbone=ViT-E/14+2024.1297.188.574.267.374.191.883.9———
E5-V (LLaVA-Next-8B)Params (B)=8.36, Backbone=LLaVA-Next-8B2024.1296.789.585.37570.189.283.5———
OpenCLIP (ViT-bigG/14)Params (B)=2.54, Backbone=ViT-bigG/142024.1296.687.974.962.575.292.284.5———
BLIP (ViT-L)Params (B)=0.23, Backbone=ViT-L2024.1296.581.769.166.676.89285.1———
EVA-CLIP (8B)Params (B)=8.222024.1296.486.674.866.174.691.382———
OpenCLIP (ViT-G/14)Params (B)=1.37, Backbone=ViT-G/142024.1295.88572.46371.291.582.1———
E5-V (LLaVA-1.5-7B)Params (B)=7.06, Backbone=LLaVA-1.5-7B2024.1295.886.681.662.96493.588———
CLIP (ViT-L)Params (B)=0.43, Backbone=ViT-L2024.1294.179.265.260.262.378.371.5———
NegCLIPParams (B)=0.152024.1292.785.976.575.275.488.882.8———
CLIP-PGS0.3Relative Training Time=0.6×2025.0388.17667.964.166.574.269.975.570.867.9
LLaVA-1.5-7BParams (B)=7.062024.128881.676.160.958.86762.4———
E-CLIPRelative Training Time=0.6x2025.0386.973.560.259.463.473.366.873.268.460.2
A-CLIPRelative Training Time=1.1x2025.0386.675.563.252.463.171.666.871.668.463.2
CLIP-PGS0.5Relative Training Time=0.5×2025.03867764.663.365.577.369.875.570.864.6
CLIPRelative Training Time=1.0×2025.0385.879.264.561.858.774.268.473.768.864.5
FLIPRelative Training Time=0.5×2025.0384.175.96660.261.671.763.27266.966