SLAKE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
SLAKE Closed
92Exact Match
11
SLAKE English
89.9Closed-Ended Accuracy
9
SLAKE Open
76.6AUROC
9
SLAKE n=350 (test)
86.3LLM Judge Accuracy
7
SLAKE Closed
90Token Recall
7
SLAKE
32.62Accuracy
6
SLAKE cross-domain
74.6Accuracy
6
SLAKE (held-out)
74.5Accuracy
6
SLAKE X-ray Patient Movement 1.0 (test)
0.875Accuracy
5
SLAKE CT Low Dose 1.0 (test)
69.7Accuracy
5
SLAKE CT Sparse View 1.0 (test)
63.64Accuracy
5
SLAKE MRI Banding 1.0 (test)
77Accuracy
5
SLAKE MRI Aliasing 1.0 (test)
60.26Accuracy
5
SLAKE MRI Human Motion 1.0 (test)
75.41Accuracy
5
SLAKE Unrelated Sentences 1.0 (test)
70.85ACC
5
SLAKE Replace Characters 1.0 (test)
71.59Accuracy
5
SLAKE Swap Characters 1.0 (test)
69.93Accuracy
5
SLAKE Add Characters 1.0 (test)
73.8Accuracy
5
SLAKE Delete Characters 1.0 (test)
73.25Accuracy
5
SLAKE-VQA
0.929COMET Score
4
SLAKE
63.8Accuracy
4
SLAKE
82.62Pass@1
4
SLAKE
0.539Final Score
4
Slake VQA
89.3Token F1
2
SLAKE X-ray Original VQA 1.0 (test)
88.89Accuracy
1