This version adds Step-aware Preference Optimization (SPO) LORA, [a novel post-training approach that independently evaluates and adjusts the denoising performance at each step, using a step-aware preference model and a step-wise resampler to ensure accurate step-aware supervision... blah, blah, blah = makes the output way artfully beautiful, a.k.a. - NICER !!!]

P.S.: The SPO approach is better than Direct Preference Optimization (DPO) that has extended its success from aligning large language models (LLMs) to aligning text-to-image diffusion models with human preferences. Unlike most existing DPO methods, which assume that all diffusion steps share a consistent preference order with the final generated images, this assumption neglects step-specific denoising performance and that preference labels should be tailored to each step's contribution.