Tag
Proposes FATE, a frame-level audio-visual temporal embedding method that aligns frame sequences on a physical timeline, enabling joint semantic and temporal understanding. It outperforms baselines on temporal retrieval, event localization, and generation evaluation metrics.