Topic metadata captured from post-publish artifact.
CVPR 2026最新研究显示自动驾驶与视频理解领域正经历从单纯内容生成向可控现实交互的范式转移，这