VIOLA: Towards Video In-Context Learning with Minimal Annotations | Signal Canvas | ScienceToStartup