AMBER
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
AMBER
24.5CHAIR
267
AMBER Generative
70.4Coverage (%)
81
AMBER
10.6CHAIR_s
56
AMBER (test)
5.6CHAIR
44
AMBER (test)
86.8Accuracy
42
AMBER
12.1CHAIR
38
AMBER
16.2CHAIR_I
35
AMBER sampled 5k
85.99A-ROC
30
AMBER-g
2.2CHAIR Score
29
AMBER
9.2CHAIR
28
AMBER Generative Task
67.1Coverage
26
AMBER Relation
81.25Accuracy
25
AMBER
8CHAIR
24
AMBER-d
89.5F1 Score
23
AMBER Discriminative Task
87.4F1 Score
22
AMBER generative subset
10.9CHAIR
22
AMBER Discriminative
89Accuracy
18
AMBER Generative
6.3CHAIRi
18
AMBER (test)
7.9CHAIR Score
18
AMBER
84.3Accuracy
18
AMBER
99.99AUC
18
AMBER
90.79Score
14
AMBER
2.98PPL
14
AMBER-d
89.2Accuracy
12
AMBER
42.3Hallucination Score (Hal)
11