⬢github Python · 775 ★ +1 since we first saw it · pushed 8 d ago · Apache-2.0
Robbyant/lingbot-touch
Learning Heterogeneous Tactile Representation for General Dexterous Manipulation
LingBot-Touch is a JEPA-based tactile representation model for robotics from Robbyant. It learns a unified latent space from heterogeneous tactile sensors (magnetic, piezoresistive, visuotactile) via modality-specific tokenizers and a shared Transformer with 5D RoPE, enabling cross-sensor transfer to real-world dexterous manipulation and high-frequency inference. The repo releases inference code and pretrained Tiny/Small checkpoints on Hugging Face and ModelScope.
Why now: Fresh release (created days ago) with open checkpoints and a paper, now trending as one of the most-starred new repos.
Who it is for: Robotics and tactile-sensing researchers building dexterous manipulation or tactile perception systems.
Stars over our 34 snapshots: 774 to 775, since 8 h ago.
Where people talked about it
- ⬢github new repos, most starred 9 min ago
API: https://socialmediatrends-api.osmike.com/v1/repos/Robbyant/lingbot-touch