DuoMatching：联合-边缘分布匹配的少步视频生成
原标题：Paper page - DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation
AI 摘要
该论文提出 DuoMatching，一种用于少步视频生成的分布匹配框架，在现有联合分布匹配基础上引入来自图像生成器的帧级边缘匹配监督，并设计 LatentBridge 解决视频学生模型与图像教师模型之间的潜表示不匹配问题，Latent Variation Sampling 则将帧级监督分散到不同时间段以减少冗余。实验表明该方法提升了视觉质量、构图和语义对齐，同时基本保持运动动态，人工评估中相对所有基线偏好率超过 80%，且不增加额外推理计算，代码已开源。
正文节选
DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation Abstract Streaming video generation has benefited from distribution matching distillation (DMD), which matches the joint distribution of video frames to a video teacher's approximation of the real video distribution. Although this joint matching mitigates drift during autoregressive rollouts, limitations remain in visual quality and semantic alignment. To address these limitations, we propose DuoMatching, a distribut