Waterbirds
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Waterbirds 100%
79.7Worst Group Accuracy
11
Waterbirds 95%
89.7Worst Group Accuracy
11
Waterbirds
92.4Worst-group Accuracy
10
Waterbirds original unshifted
90.8Worst Accuracy
10
Waterbirds CB
77.9Accuracy
10
Waterbirds (test)
63.64Avg Acc (0.5% Bias)
10
Waterbirds 5.0 severity (test)
66.33Accuracy
10
Waterbirds 2.0 severity (test)
65.23Accuracy
10
Waterbirds severity 1.0 (test)
0.652Accuracy
10
Waterbirds severity 0.5 (test)
63.64Accuracy
10
Waterbirds
28.2Delta WGA
9
Waterbirds
92.26Top-1 Group Accuracy
9
Waterbirds 5.0% bias-conflict ratio
81.22Bias Conflict (BC)
9
Waterbirds 1.0% bias-conflict ratio
70.7Bias Conflict (BC)
9
Waterbirds (OOD)
81.1Accuracy
9
Waterbirds (ID)
81.7Accuracy
9
Waterbirds (target domain)
93.13Avg Acc
9
Waterbirds (test)
15.5Balanced Error
9
Waterbirds standard (test)
100Precision@10
8
Waterbirds
90.5OOD Accuracy
8
WaterBirds biased scenario (test)
87.42Average Accuracy
7
Waterbirds 100-reverse
93Mean Accuracy
6
Waterbirds 100
89.5Mean Accuracy
6
Waterbirds
0Concept Leakage Proportion
6
Waterbirds
69.18Worst-Group Acc
5