Cycle Consistency as Reward: Learning Image-Text Alignment Without Human Preferences.

Hyojin Bahng, Caroline Chan, Frdo Durand, Phillip Isola

Browse the full ICCV paper archive.