ReLAP pretrains robot actions from video via recurrent prediction in JEPA space
A new arXiv paper, ReLAP, proposes latent action pretraining through recurrent video prediction in JEPA space, aiming to learn robot behavior from unlabeled videos without action annotations.
Read the full story at arXiv cs.RO ↗
Timeline · 1 report
- 2026-11-18 05:00 · arXiv cs.RO
ReLAP: Latent Action Pretraining through Recurrent Video Prediction in JEPA Space