Waterbirds
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Waterbirds
97.2WG Accuracy
283
Waterbirds (test)
97.1Worst-Group Accuracy
214
Waterbirds In-context training: iNaturalist (test)
92.23Average Worst-Group Accuracy
140
Waterbirds severe in-distribution context (test)
91.38Worst-group Accuracy
104
Waterbirds In-context training (test)
93.4Average Worst-Group Accuracy
104
Waterbirds (test)
98.2Test Accuracy
40
Waterbirds Background (test)
91.8Accuracy
24
Waterbirds Background
88.6Accuracy (Zero-shot)
24
Waterbirds 95% correlation (test)
92.5Worst-group Accuracy
23
Waterbirds (test)
92.9WGA
22
Waterbirds 100%
16Accuracy Variance across Groups
22
Waterbirds 95%
6Accuracy Variance (Group)
22
Waterbirds (test)
90Worst-Group Accuracy
22
Waterbirds 100% correlation (test)
91.9Worst-group Accuracy
21
Waterbirds
16.68Relative Accuracy Improvement
18
Waterbirds (OOD)
98.9Accuracy
18
Waterbirds 95% (test)
91.8Worst-Group Accuracy
18
Waterbirds
83Precision@10
17
Waterbirds 100% (test)
90.7Worst Group Accuracy
17
WaterBirds (test)
98.5Worst Group Accuracy
15
Waterbirds Flip (test)
90.6Accuracy
14
Waterbirds Original (test)
98.4Accuracy
14
Waterbirds (test)
92.67WGA
13
Waterbirds
92.12Average Accuracy (AA)
12
Waterbirds bias-free (test)
90.8Balanced Accuracy
11