POPE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
POPE v1.0 (test)
87POPE Score
15
POPE MSCOCO, A-OKVQA, GQA average (Adversarial)
84.11Accuracy
15
POPE Popular MSCOCO, A-OKVQA, GQA average
88.48Accuracy
15
POPE MSCOCO, A-OKVQA, GQA average (Random)
93.24Accuracy
15
POPE v2
100Accuracy
15
POPE GQA 2019 (Random)
89.3Accuracy
15
POPE MSCOCO (test)
87.51Random Accuracy
15
POPE
85.56Score
14
POPE
89Accuracy
14
POPE (average)
88.93F1 Score
14
POPE (dev)
87Accuracy
13
POPE COCO 2014 (Adversarial)
81.43Accuracy
13
POPE COCO 2014 (Random)
90.1Accuracy
13
POPE 2023
89.9Accuracy
13
POPE MS-COCO Overall
86.83Accuracy
12
POPE Exact
100Success Rate
12
POPE Random
92.1Accuracy
12
POPE averaged across MS-COCO, A-OKVQA, and GQA (Adversarial)
0.807Accuracy
12
POPE averaged across MS-COCO, A-OKVQA, and GQA (Popular)
85.2Accuracy
12
POPE averaged across MS-COCO, A-OKVQA, and GQA (Random)
90.1Accuracy
12
POPE (test)
87.9POPE Score
12
POPE
89.1Accuracy
11
POPE Random
89.87Accuracy
11
POPE A-OKVQA 2022 (Adversarial)
80.13Accuracy
11
POPE MS COCO v1 (test)
86Random Accuracy
11