VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection Paper • 2411.14794 • Published 11 days ago • 10