Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model | Read Paper on Bytez