Text-to-Image Generation on ImageNet 256x256 (OOD)
3.23FIDComposer
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ComposerCategory=VAR d-16, Step=10, Parameter=412M, Time=0.42 (+5%), Memory=2.57G2026.03 | 3.23 | 281.4 | 87 | 56 | |
| Test-time TrainingCategory=VAR d-16, Step=10, Parameter=310M, Time=40.52 (+10,030%), Memory=4.58G2026.03 | 3.28 | 279.1 | 85 | 53 |