MetaFold was proposed by a joint team from the National University of Singapore and other universities in August 2025. The relevant research results are "MetaFold: Language-Guided Multi-Category Garment Folding Framework via Trajectory Generation and Foundation Model", has been IROS25 accept.

MetaFold is a unified framework that decouples garment folding task planning from motion prediction, learning each component independently to enhance model generalization. It employs language-guided point cloud trajectory generation for task planning and a low-level base model for motion prediction. This architecture facilitates multi-class learning, enabling the model to flexibly adapt to a variety of user commands and folding tasks.