List your company

Knowledge-based Visual Retrieval on OKVQA Google Search (test)

84.66PR@5MIRe
32.535246.067659.673.1324Nov 13, 2024

Evaluation results

MethodLinks
MIReMode=Fine-tuning, Backbone=ViT-large2024.1184.66—91.3
MIReMode=Fine-tuning2024.1183.59—90.59
VISTAMode=Fine-tuning2024.1182.06—90.11
MIReZero-shot=true, Backbone=ViT-large2024.1181.1363.1788.72
MIReZero-shot=true2024.1180.4863.0388.15
GeMKRMode=Fine-tuning2024.1178.6—86.2
PreFLMR†Zero-shot=true, Pre-training=trained on same dataset and experimental settings as current paper2024.1176.8359.3884.34
MIReMode=Fine-tuning, Pre-training=None2024.1174.26—84.07
UniIRZero-shot=true2024.1173.9453.2784.19
ReVizMode=Fine-tuning2024.1173.35—83.17
VISTAZero-shot=true2024.1172.8355.3381.61
VRRMode=Fine-tuning2024.1171.5—81.5
FLMRMode=Fine-tuning2024.1170.63—81.23
ReVizZero-shot=true2024.1164.0545.7775.39
FLMRZero-shot=true2024.1157.2538.1569.42
CLIPZero-shot=true2024.1134.5419.0850.48