arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.05871cs.CV
收藏

逐帧测试时校正用于自回归长视频生成

Pathwise Test-Time Correction for Autoregressive Long Video Generation

Nanjing University(南京大学);Tencent Hunyuan(腾讯文言);Chinese University of Hong Kong Shenzhen(香港中文大学(深圳));University of Science and Technology of China(中国科学技术大学)

Xunzhi Xiang, Zixuan Duan, Guiyu Zhang, Haiyu Zhang, Zhe Gao, Junta Wu, Shaofeng Zhang, Tengfei Wang, Qi Fan, Chunchao Guo

AI总结:

本文提出TTC方法,通过利用初始帧作为参考锚点,有效校正自回归长视频生成中的漂移问题,提升生成质量并延长生成长度。

摘要:

压缩的自回归扩散模型有助于实时短视频合成,但在长序列生成过程中会出现严重的误差累积。尽管现有的测试时间优化(TTO)方法在图像或短片段上效果显著,但我们发现它们无法缓解扩展序列中的漂移,因为奖励景观不稳定且压缩参数高度敏感。为克服这些限制,我们引入了测试时间校正(TTC),一种无需训练的替代方法。具体而言,TTC利用初始帧作为稳定的参考锚点,以校准采样轨迹上的中间随机状态。广泛的实验表明,我们的方法能够无缝集成到各种压缩模型中,扩展生成长度的同时几乎没有开销,并在30秒基准上与资源密集型训练方法的性能相当。

英文摘要:

Distilled autoregressive diffusion models facilitate real-time short video synthesis but suffer from severe error accumulation during long-sequence generation. While existing Test-Time Optimization (TTO) methods prove effective for images or short clips, we identify that they fail to mitigate drift in extended sequences due to unstable reward landscapes and the hypersensitivity of distilled parameters. To overcome these limitations, we introduce Test-Time Correction (TTC), a training-free alternative. Specifically, TTC utilizes the initial frame as a stable reference anchor to calibrate intermediate stochastic states along the sampling trajectory. Extensive experiments demonstrate that our method seamlessly integrates with various distilled models, extending generation lengths with negligible overhead while matching the quality of resource-intensive training-based methods on 30-second benchmarks.

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。

↑