hpsv3-human-preference-score-evaluation

【效率与协作】A VLM-based preference scoring system trained on 1.17M annotated comparisons to evaluate text-to-image generation quality at scale. Uses uncertainty-aware ranking loss for fine-grained assessment across diverse images and supports iterative quality improvement through chain-of-human-preference sampling.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准