Compositional text-to-image generation
Benchmarks
DatasetSOTA methodMetricTrendResultsLast updated
85.08Color
17
Apr 7, 2026
0.794Color Accuracy
16
May 28, 2026
100Single Object Accuracy
13
May 29, 2026
86.92Color Score
13
Mar 20, 2026
0.322CLIP-Score
8
Apr 29, 2026
85.11Accuracy
8
Feb 26, 2026
27.9CLIP Score
6
Apr 29, 2026
28.9CLIP Score
6
Apr 29, 2026
0.31CLIP Score
6
Apr 29, 2026
0.275CLIP Score
6
Apr 29, 2026
56.29SG-IoU
4
May 12, 2026
52.2Atomicity Score (N=3)
3
May 28, 2026
0.382BLIP-VQA Score (Color)
2
Jun 12, 2026