Assessing AI systems across multiple input/output types (audio, video, text) simultaneously rather than separately.