视频生成只学“接得上”，没学“会演变”;这篇补上了
现在的 AI 生成视频是一段一段往外吐的，训练时只要求“下一段接得上上一段”。但接得上不等于懂物理：画面里球滚过去，下一帧位置对，不代表模型真在推演运动，它可能只是靠相邻帧长得像来蒙混过关。研究者加了一个辅助任务：让模型在看不到中间画面的情况下，直接预测隔开一段时间的未来片段——逼它必须理解“场景怎么演变”，而不是抄近道。结果物理合理性更好，训练成本反而更低。它不是你明天能用上的功能，但这是视频生成从“会接”走向“会想”的一个关键信号。
📄 原文摘要(英文)
Autoregressive video diffusion models generate videos chunk by chunk, yet producing plausible continuations does not guarantee that they capture how scenes evolve. Under teacher forcing, strong temporal redundancy may allow local visual continuity to serve as a shortcut for current-chunk denoising, reducing the pressure to model visual evolution. We introduce Future Chunk Prediction (FCP), an auxiliary objective jointly optimized with teacher-forced denoising that predicts temporally separated chunks without access to the intervening visual context. Drawing supervision directly from training videos, FCP requires no additional pretrained visual encoder and leaves autoregressive inference unchanged. Our method achieves superior physical plausibility over standard teacher-forced autoregressive training with a lower overall training compute budget, highlighting the effectiveness and efficiency of future-chunk supervision for modeling visual evolution. Intriguingly, our visualizations suggest that FCP fosters a more structured representation space, which may be valuable for building autoregressive world models.