POPE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
POPE MSCOCO
88.47Accuracy
24
POPE Adv
94Precision
24
POPE MSCOCO (test)
86.15Average F1
24
POPE GQA (random)
92.23Accuracy
24
POPE Popular v1.0
88.03Accuracy
24
POPE v1.0 (Random)
90.07Accuracy
24
POPE Average of Random, Popular, and Adversarial 2023
89.4Accuracy
24
POPE
89.1POPE Score
22
POPE average across COCO, A-OKVQA, GQA
85.7ACC
22
POPE
86.4Accuracy (POPE)
21
POPE average (MS-COCO, A-OKVQA, GQA)
89.21Accuracy
21
POPE Overall
91.1Accuracy
21
POPE Adversarial
97.1AUROC
21
POPE MSCOCO (val)
88.1F1 Score
21
POPE
90.4POPE Score
20
POPE
88.41Accuracy
18
POPE
86.73Average F1
18
POPE Concurrence
92.33Accuracy
18
POPE 28
88.9Accuracy
18
POPE
93.53Ins Score
16
POPE
85.9F1 Score
16
POPE pt-PT
89.6Accuracy
16
POPE adversarial (400-sample held-out set)
89.3Accuracy
16
POPE
84.7Accuracy
16
POPE average across three subsets
88.93Accuracy
16