注意生成细节：面向视频扩散模型的直接局部化细节偏好优化
Mind the Generative Details: Direct Localized Detail Preference Optimization for Video Diffusion Models
AI总结：
本文提出LocalDPO，一种新的后训练框架，通过从真实视频中构建局部偏好对，并在时空区域层面优化对齐，以提高视频生成的质量和人类偏好评分。
摘要：
将文本到视频的扩散模型与人类偏好对齐对于生成高质量视频至关重要。现有的直接偏好优化（DPO）方法依赖于多样本排序和任务特定的批评模型，这效率低下且常导致模糊的全局监督。为了解决这些限制，我们提出了LocalDPO，一种新的后训练框架，该框架从真实视频中构建局部偏好对，并在时空区域层面进行优化。我们设计了一个自动化流程，高效地收集偏好对数据，通过单次提示推理生成偏好对，消除了对外部批评模型或人工标注的需求。具体来说，我们将高质量的真实视频作为正样本，并通过局部随机时空掩码来生成对应的负样本，仅使用冻结的基模型恢复被掩码的区域。在训练过程中，我们引入了区域感知的DPO损失，将偏好学习限制在被损坏的区域以实现快速收敛。在Wan2.1和CogVideoX上的实验表明，LocalDPO在视频保真度、时间连贯性和人类偏好评分方面优于其他后训练方法，建立了更高效和精细的视频生成器对齐范式。代码可在https://github.com/1170300714/Local-DPO上获得。
英文摘要：
Aligning text-to-video diffusion models with human preferences is crucial for generating high-quality videos. Existing Direct Preference Otimization (DPO) methods rely on multi-sample ranking and task-specific critic models, which is inefficient and often yields ambiguous global supervision. To address these limitations, we propose LocalDPO, a novel post-training framework that constructs localized preference pairs from real videos and optimizes alignment at the spatio-temporal region level. We design an automated pipeline to efficiently collect preference pair data that generates preference pairs with a single inference per prompt, eliminating the need for external critic models or manual annotation. Specifically, we treat high-quality real videos as positive samples and generate corresponding negatives by locally corrupting them with random spatio-temporal masks and restoring only the masked regions using the frozen base model. During training, we introduce a region-aware DPO loss that restricts preference learning to corrupted areas for rapid convergence. Experiments on Wan2.1 and CogVideoX demonstrate that LocalDPO consistently improves video fidelity, temporal coherence and human preference scores over other post-training approaches, establishing a more efficient and fine-grained paradigm for video generator alignment.The code is available at https://github.com/1170300714/Local-DPO.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
