WM-Craftnet: World Synesthesia Model for Generalizable and Robust Dexterous In-Hand Manipulation
Classification
View four quadrants- Major category
- WAMs
- Quadrant
- Outside quadrants
- Architecture
- Pending verification
- Prediction paradigm
- Other mechanisms
- Source review status
- Not assigned
Category review. An action-conditioned recurrent world model predicts visuotactile state and conditions a separate closed-loop dexterous policy. This is task-specific world-model-conditioned control despite no imagined policy rollouts. The described action mechanism is external planning, model-assisted policy optimization, geometric tracking, or video-conditioned control; the source does not establish joint future/action generation or an IDM action decoder. Reading evidence
Contribution
WM-Craftnet learns a predictive visuotactile state that conditions a separate dexterous manipulation policy. Its main contribution is recurrent perception for executed control: noisy depth, touch, proprioception, and action history inform a latent state trained with clean-depth and other prediction targets. Hardware rotation improves substantially, but severe-disturbance recovery remains limited (e-rssm, e-policy, e-stress).
Abstract
An abstract has not been added yet.
Affiliations
Sharpa Robotics