Applying attention consistency training to omni-modal temporal grounding is a first, though similar ideas have precedents in visual grounding and interpretability research.