GQA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
GQA
62.5GQA Score
7
GQA (dev)
63.8Accuracy
7
GQA
61.3Exact Match
7
GQA (Orig.)
58.35Overall Accuracy
7
GQA SGL compositional generalization (test)
61.4Accuracy
6
GQA Agentic
57.7Accuracy
6
GQA Spatial
31.6Accuracy
6
GQA 200
34.1R@50
6
GQA 200
63.8R@50
6
GQA ZOOM-Hard
51.44GPT Accuracy
6
GQA ZOOM-Medium
59.31GPT Accuracy
6
GQA ZOOM-Easy
68.75GPT Accuracy
6
GQA-OOD (val)
78Accuracy
6
GQA-REX
61.4BLEU-4
6
GQA
61.1Score
6
GQA balanced (test-dev)
58.9Accuracy
6
GQA
1,699Token Count
6
GQA (test)
30.2R@50
6
GQA (test)
65R@50
6
GQA
71.2Top-1 Accuracy
5
GQA
0.814Kendall τ
5
GQA (testdev)
61.03Overall Accuracy
5
GQA isolated (test)
100Accuracy
5
GQA connected (test)
100Accuracy
5
GQA in_between (test)
99.23Accuracy
5