解决长视频生成中误差累积问题，无需训练即可稳定输出。
Pathwise Test-Time Correction for Autoregressive Long Video Generation
- 以首帧为参考锚点，动态校准生成过程中的中间状态。
- 在30秒视频上实现高质量生成，性能媲美耗时训练方法。
- 适合作为现有模型的轻量级插件，提升长序列生成稳定性。
蒸馏版自回归扩散模型可实现实时短视频生成，但在长序列生成中存在严重误差累积问题。现有测试时优化（TTO）方法虽对图像或短片段有效，但因奖励景观不稳定及蒸馏参数敏感，难以缓解长序列漂移。为此，我们提出测试时校正（TTC），一种无需训练的替代方案。TTC利用初始帧作为稳定参考锚点，对采样轨迹上的中间随机状态进行校准。大量实验表明，该方法可无缝集成至多种蒸馏模型，在几乎无额外开销下扩展生成长度，且在30秒基准测试上达到与资源密集型训练方法相当的质量。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。