Event ·Single source

ViGeo:用时空类比把视觉上下文学习扩展到视频任务

研究团队提出 ViGeo 框架,通过时空画布补全(spatiotemporal canvas completion)把视觉类比式上下文学习从图像扩展到视频。该方法免训练即可适配多种视频任务,并能零样本泛化到未见过的视频操作和 event cameras 等新模态。论文还发现任务内化现象:预训练任务的查询格式会覆盖示例,加入少量任务无关数据即可消除这一捷径。

Current read

训练免了!ViGeo 用时空类比让视频模型靠示例学新任务,还能零样本处理 event cameras,顺手拆了个提示词捷径的坑。

2 sources43° AI heatLast updated 2026/9/27 21:26:58

Evidence

2 sources

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.