推测解码用于自回归视频生成
Speculative Decoding for Autoregressive Video Generation
AI总结：
本文探讨了推测解码在自回归视频生成中的应用，通过引入SDVG框架，利用图像质量路由替代token验证，实现高效视频生成，同时保持高质量并提升生成速度。
摘要：
自回归视频扩散正成为视频合成的有希望范式，步蒸馏是加速推理的主要方法。推测解码，大型语言模型的主要加速策略，能否有效适应自回归视频生成仍是一个开放问题，因为视频块是连续的时空张量，没有token级分布用于精确拒绝采样。我们引入SDVG，通过将token验证替换为图像质量路由，将推测解码引入基于块的自回归视频扩散中。一个13亿参数的drafter通过四步去噪生成候选块；每个块通过VAE解码并由ImageReward评分，采用最差帧聚合——取每帧评分的最小值以捕捉单帧伪影，平均会掩盖。评分高于固定阈值tau的块被接受进入14B目标的KV缓存；其余块由目标重新生成。两个额外的设计选择证明关键：第一个块始终被强制拒绝以锚定场景组成，tau作为单一调节器，追踪平滑的质量-速度帕累托前沿。在1003个MovieGenVideoBench提示（832x480）上，SDVG在tau=-0.7时以1.59倍速度提升，保留98.1%的目标仅VisionReward质量（0.0773 vs. 0.0788），并在95.7%质量保留时达到2.09倍速度——同时始终优于仅draft生成的+17%以上。该框架无需训练，不需要架构更改，可无缝集成到现有自回归视频生成管道中。
英文摘要：
Autoregressive video diffusion is emerging as a promising paradigm for streaming video synthesis, with step distillation serving as the primary means of accelerating inference. Whether speculative decoding, the dominant acceleration strategy for large language models, can be effectively adapted to autoregressive video generation remains an open question, because video blocks are continuous spatiotemporal tensors with no token-level distribution for exact rejection sampling. We introduce SDVG, which brings speculative decoding to block-based autoregressive video diffusion by replacing token verification with an image-quality router. A 1.3B drafter proposes candidate blocks via four denoising steps; each block is VAE-decoded and scored by ImageReward using worst-frame aggregation--taking the minimum per-frame reward to catch single-frame artifacts that averaging would mask. Blocks scoring above a fixed threshold tau are accepted into the 14B target's KV cache; the rest are regenerated by the target. Two additional design choices prove critical: the first block is always force-rejected to anchor scene composition, and tau serves as a single knob that traces a smooth quality-speed Pareto frontier. On 1003 MovieGenVideoBench prompts (832x480), SDVG retains 98.1% of target-only VisionReward quality (0.0773 vs. 0.0788) at a 1.59x speedup with tau=-0.7, and reaches 2.09x at 95.7% quality retention--while consistently outperforming draft-only generation by over +17%. The framework is training-free, requires no architectural changes, and can be seamlessly integrated into existing autoregressive video generation pipelines.
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。
