Universal Video Temporal Grounding with Generative Multi-modal Large Language Models

This paper presents a computational model for universal video temporal grounding, which accurately localizes temporal moments in videos based on natural language queries (e.g., questions or descriptions).