aiqbench analysis
Instant-Fold advances robotic manipulation of deformable objects by leveraging in-context imitation learning and flow-matching transformers to generalize from single demonstrations without gradient updates. This work addresses the high-dimensional state spaces and topology-changing interactions inherent in folding tasks, demonstrating zero-shot transfer from simulation to real-world environments. It represents a significant methodological step in bridging computer vision and robotics for complex physical interactions.
Instant-Fold: In-Context Imitation Learning for Deformable Object Manipulation
arXiv:2606.04269v2 Announce Type: replace-cross Abstract: Deformable object manipulation (DOM) is challenging due to high-dimensional, partially observable states that evolve through long-horizon, topology-changing interactions with multiple valid manipulation modes. We introduce Instant-Fold, an in-context imitation learning framework for DOM. Given a single human demonstration, our policy infers and executes diverse manipulation modes directly from the demonstration, including variations in spatial execution and ordering, without requiring gradient updates. Our approach first learns deformation-aware visual representations via temporal contrastive pretraining, after which a flow-matching transformer policy conditioned on the demonstration predicts actions to execute the intended manipulation mode. Trained entirely in simulation, Instant-Fold generalizes across diverse folding modes and transfers zero-shot to real-world settings without additional data collection or finetuning. Videos are available at https://instant-fold.github.io.
Abstract:Deformable object manipulation (DOM) is challenging due to high-dimensional, partially observable states that evolve through long-horizon, topology-changing interactions with multiple valid manipulation modes. We introduce Instant-Fold, an in-context imitation learning framework for DOM. Given a single human demonstration, our policy infers and executes diverse manipulation modes directly from the demonstration, including variations in spatial execution and ordering, without requiring gradient updates. Our approach first learns deformation-aware visual representations via temporal contrastive pretraining, after which a flow-matching transformer policy conditioned on the demonstration predicts actions to execute the intended manipulation mode. Trained entirely in simulation, Instant-Fold generalizes across diverse folding modes and transfers zero-shot to real-world settings without additional data collection or finetuning. Videos are available at this https URL.
Accepted at CoRL 2026. Project Webpage: this https URL
Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2606.04269 [cs.RO]
(or arXiv:2606.04269v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2606.04269 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 228 words
Related coverage
Instant-Fold: In-Context Imitation Learning for Deformable Object Manipulation
arXiv · Jun 4, 2026
Game Design Creativity evals
LessWrong · Oct 12, 2026
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
Touch2Trace: Tactile-Driven Imitation Learning for Dexterous Cable Tracing
arXiv · Sep 15, 2026