Compositional Reasoning
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
95.2Overall Accuracy
95
Jul 7, 2026
89.2Average Score
65
Jun 26, 2026
81.8Attribute Score
47
Jun 12, 2026
41.25Group Score
33
Jul 7, 2026
75.2Average Performance
25
May 13, 2026
93.1Sugarcrepe Score
23
Apr 10, 2026
44.9Accuracy
21
Jun 26, 2026
51.3Accuracy
21
Jun 26, 2026
70.8Level-1 Score
21
Apr 30, 2026
62Accuracy (GQA Compositional Reasoning)
18
Jun 23, 2026
83.79Spatial Score
18
Mar 12, 2026
83.6Relation Score
17
Jun 4, 2026
94.6Accuracy
16
Jul 7, 2026
94.88Text Score
16
Jun 23, 2026
64.7Accuracy
16
Apr 14, 2026
57.7Accuracy
13
Jul 7, 2026
68Accuracy
12
Apr 14, 2026
23.1Accuracy
11
Feb 26, 2026
48.9Accuracy
11
Feb 26, 2026
33.6Accuracy
11
Feb 26, 2026
44.28Text Attribute Accuracy
11
Feb 26, 2026
0.67Text Score
11
Feb 26, 2026
35.5Accuracy
10
Mar 19, 2026
33.33Txt2Img Score
10
Feb 26, 2026
60Logical Question Accuracy
8
Jun 23, 2026