We include several videos here: two clips each of agents performing the 2D and 3D odd-one-out tasks, and a higher-resolution video of a human performing the 3D task.
