FlashMotion:基于轨迹引导的少步可控视频生成
TLDR
FlashMotion 提出一种训练框架,将轨迹可控视频生成器蒸馏为少步模型,并推出 FlashBench,提升质量与轨迹准确性。
评分理由
The paper addresses computational inefficiency in multi-step trajectory-controllable video generation with a distillation and adapter finetuning strategy, and contributes a benchmark. However, the abstract does not provide full experimental details or limitations, and it does not frame the work as a world model.
Read-first 评分解释
综合优先阅读分 30.8,由主题、引用、图谱、方法、可复现性和近期性等信号加权得到。 原始总分保留为 0。
研究版图角色
前沿论文
排序敏感性
稳定性:volatile;排名波动范围:607。