Jinglianwen adds force and tactile signals to its robot-training corpus
Jinglianwen Technology released a synchronized training dataset combining robot actions, RGB and RGB-D vision, audio, object state and high-density tactile signals.
Evidence notes
- The dataset records joint state, end-effector pose and 3,750 tactile-pressure channels per hand.
- It aligns visual, force, tactile, audio and object-state signals at millisecond scale.
- Jinglianwen describes use in VLA training, dexterous manipulation, force control, behaviour cloning and reinforcement learning.
Company context
AI data infrastructure company providing multimodal collection, governance, annotation, quality control and dataset operations for model developers, robotics teams and public-sector data programmes.