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