Causal Forcing++：用于实时交互式视频生成的可扩展少步自回归扩散蒸馏
Causal Forcing++: Scalable Few-Step Autoregressive Diffusion Distillation for Real-Time Interactive Video Generation
AI总结：
提出Causal Forcing++框架，通过因果一致性蒸馏（causal CD）实现帧级1-2步自回归扩散蒸馏，在降低延迟和训练成本的同时提升视频生成质量。
摘要：
实时交互式视频生成需要低延迟、流式处理和可控展开。现有的自回归（AR）扩散蒸馏方法通过将双向基础模型蒸馏为少步AR学生模型，在分块4步机制中取得了强劲结果，但仍受限于粗粒度响应和不可忽略的采样延迟。本文研究了一种更激进的设置：仅用1-2采样步的帧级自回归。在此机制下，我们识别出少步AR学生模型的初始化是关键瓶颈：现有策略要么目标不对齐，要么无法进行少步生成，要么成本过高难以扩展。我们提出 extbf{Causal Forcing++}，一个原则性且可扩展的流水线，使用\emph{因果一致性蒸馏}（causal CD）进行少步AR初始化。核心思想是：因果CD学习与因果ODE蒸馏相同的AR条件流映射，但通过相邻时间步之间的单个在线教师ODE步获得监督，避免了预计算和存储完整PF-ODE轨迹的需要。这使得初始化既更高效又更易优化。由此产生的流水线\ours在 extit{ extbf{帧级2步设置}}下，VBench总分、VBench质量和VisionReward分别超过SOTA 4步分块Causal Forcing 0.1、0.3和0.335，同时首帧延迟降低50%，阶段2训练成本降低约$4 imes$。我们进一步将流水线扩展到以动作条件的世界模型生成，秉承Genie3的精神。项目页面：https://github.com/thu-ml/Causal-Forcing 和 https://github.com/shengshu-ai/minWM 。
英文摘要：
Real-time interactive video generation requires low-latency, streaming, and controllable rollout. Existing autoregressive (AR) diffusion distillation methods have achieved strong results in the chunk-wise 4-step regime by distilling bidirectional base models into few-step AR students, but they remain limited by coarse response granularity and non-negligible sampling latency. In this paper, we study a more aggressive setting: frame-wise autoregression with only 1--2 sampling steps. In this regime, we identify the initialization of a few-step AR student as the key bottleneck: existing strategies are either target-misaligned, incapable of few-step generation, or too costly to scale. We propose \textbf{Causal Forcing++}, a principled and scalable pipeline that uses \emph{causal consistency distillation} (causal CD) for few-step AR initialization. The core idea is that causal CD learns the same AR-conditional flow map as causal ODE distillation, but obtains supervision from a single online teacher ODE step between adjacent timesteps, avoiding the need to precompute and store full PF-ODE trajectories. This makes the initialization both more efficient and easier to optimize. The resulting pipeline, \ours, surpasses the SOTA 4-step chunk-wise Causal Forcing under the \textit{\textbf{frame-wise 2-step setting}} by 0.1 in VBench Total, 0.3 in VBench Quality, and 0.335 in VisionReward, while reducing first-frame latency by 50\% and Stage 2 training cost by $\sim$$4\times$. We further extend the pipeline to action-conditioned world model generation in the spirit of Genie3. Project Page: https://github.com/thu-ml/Causal-Forcing and https://github.com/shengshu-ai/minWM .
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。
