Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher

作者：Bowen Zheng, Zhiguang Liu, Jiarong Ou, Rui Chen, Tianyang Hu · 单位：Affiliation: The Chinese University of Hong Kong, Shenzhen · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-09