FOIL
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
FOIL
99.2Accuracy (4-ref)
33
FOIL
98.4Accuracy (4 Refs)
32
FOIL (test)
95.4Accuracy
28
Foil
0.245AURC
23
FOIL
92.6Accuracy
18
FOIL (4-ref)
94.64Accuracy
15
FOIL 1-ref
90.94Accuracy
15
FOIL R4
95.4P-Acc
10
FOILR1
95.1Preference Accuracy
10
FOIL 30K captions derived from COCO
84.3Object Accuracy
10
FOIL (test)
98.4Accuracy
9
FOIL 4-ref
98.4Accuracy
6
FOIL (1-ref)
98.2Accuracy
6
FOIL-nocaps (Out of Domain)
80.2FDR
6
FOIL nocaps (In Domain)
78.9FDR
6
FOIL nocaps (Overall)
18.6FDR
6