Mind the Generative Details: Direct Localized Detail Preference Optimization for Video Diffusion Models
This addresses the need for more efficient and fine-grained alignment in video generation, though it is incremental as it builds on existing Direct Preference Optimization methods.
The paper tackled the problem of aligning text-to-video diffusion models with human preferences by proposing LocalDPO, a post-training framework that constructs localized preference pairs from real videos and optimizes alignment at the spatio-temporal region level, resulting in improved video fidelity, temporal coherence, and human preference scores over other approaches on datasets like Wan2.1 and CogVideoX.
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.