FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding
FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding
FoldBack is a self-correcting masked generative policy for long-horizon garment folding that enables failed interactions to be detected, undone, and repaired before execution continues. It achieves 75.2% final folding success and 0.837 final-mask IoU across 33 real garments, outperforming prior baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
final-mask IoU
0.837
final-mask IoU
0.689
real garments
33
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
nabling failed interactions to be detected, undone, and repaired before execution continues, without recovery demonstrations or base-policy retraining. Across 33 real garments from six categories, FoldBack achieves 75.2% final folding success and 0.837 final-mask IoU, versus 45.7% and 0.689 for the strongest prior baseline.
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org