POPE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
POPE random
89.1F1 Score
6
POPE Adversarial
88.17Accuracy
6
POPE Popular
89.7Accuracy
6
POPE (val)
93.7Accuracy
6
PoPe
88.3Accuracy
6
POPE MS-COCO Popular sampling (val)
82.77Accuracy
6
POPE MS-COCO Random sampling (val)
88.57Accuracy
6
POPE
437Latency (10/token) (ms)
5
POPE Objects365
91.47F1 Score
5
POPE (test)
89.4Final Performance Score
5
POPE (test)
87.3Accuracy
4
POPE (adversarial)
73AUROC (Global)
4
POPE popular
77AUROC (Global)
4
POPE random
0.82AUROC (Global)
4
POPE
82.8Score (Noise)
4
POPE (test)
26Total Time
4
POPE
84.5Accuracy
4
POPE
96.1Accuracy
4
POPE (test)
756,000Total Time (ms)
4
POPE Adversarial LLaVA-1.5
77.67Score (Zh)
4
POPE Random LLaVA-1.5
86.5Zh Score
4
POPE
90.29Accuracy
3
POPE
77.87Adversarial Score
3
POPE Adversarial (test)
97.1AUROC
3
POPE
86.4Accuracy
3