●builderExpect current omni models to fail at precise temporal tasks like event localization or audio-visual sync in your applications.
●researcherThis provides a new gold standard for testing how well multimodal models actually synchronize audio and video events.