VideoAgentTrek: Computer Use Pretraining from Unlabeled Videos
videoagenttrek.github.io
VideoAgentTrek: Computer Use Pretraining from Unlabeled Videos - A novel approach to training computer-use agents using unlabeled video data
关键词:VideoAgentTrek, Computer-Use Agent, Video Pretraining, Unlabeled Videos, Vision-Language Model, Agent Learning, Self-Supervised Learning