arXiv Robotics — research abstracts· Jeongeun Park, Taekyung Kim, Dongyoon Han, Sangdoo Yun·· 1 days agoEditorial score65
Learning Transferable World-Action Models from Task-Paired Human Videos
Learning Transferable World-Action Models from Task-Paired Human Videos
Summary
The paper presents WATCH, a cross-video pre-training method that learns transferable world-action models from task-paired human demonstrations. By jointly predicting actions and future frames across different demonstrations, WATCH achieves improved motion alignment and better performance on unseen robots, outperforming within-video pre-training and co-training baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org