研究人员推出了 ViPO，这是一个大规模数据集，旨在通过偏好优化来改进视觉生成模型。该数据集包含 100 万张图像对和 30 万个视频对，解决了现有数据集分辨率低和分布不平衡等局限性。他们还开发了 Poly-DPO，一种增强对嘈杂偏好数据鲁棒性的算法，在现有数据集上取得了显著的提升，并与 ViPO 结合使用时表现更优。 AI
影响 通过提供大规模、高质量的偏好数据集和鲁棒的优化算法，提升了视觉生成模型的质量。
排序理由 介绍用于视觉生成模型的新数据集和优化技术的学术论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
While preference optimization is crucial for improving visual generative models, how to effectively scale this paradigm remains largely unexplored. Current open-source preference datasets contain conflicting preference patterns, where winners excel in some dimensions but underper…
arXiv:2604.24953v1 Announce Type: new Abstract: While preference optimization is crucial for improving visual generative models, how to effectively scale this paradigm remains largely unexplored. Current open-source preference datasets contain conflicting preference patterns, whe…
While preference optimization is crucial for improving visual generative models, how to effectively scale this paradigm remains largely unexplored. Current open-source preference datasets contain conflicting preference patterns, where winners excel in some dimensions but underper…