Google DeepMind Introduces Direct Reward Fine-Tuning (DRaFT): An Effective Artificial …
The authors of the study applied DRaFT to Stable Diffusion 1.4 and conducted evaluations using various reward functions and prompt sets. Their ...