X Square Robot open-sources HOST one-shot skill-acquisition framework
X Square Robot and academic collaborators open-sourced HOST, a framework that learns a manipulation skill from one short human demonstration video at inference time without task-specific fine-tuning or parameter updates.
Why it matters
X Square Robot researchers and academic collaborators released HOST, a framework that acquires new manipulation skills at inference time from a single human demonstration video without task-specific parameter updates.
Evidence notes
- HOST was evaluated across 50 novel manipulation tasks and achieved 62% average success on the eight-task baseline-comparison set.
- Average skill-acquisition time was 29 seconds, with a 43-point advantage over the strongest compared method without parameter updates.
- The project page, implementation code and paper are publicly available; the work was developed with researchers from Tsinghua University and Beijing Institute of Technology.
Company context
Full-stack embodied-AI robotics company developing WALL foundation models, Quanta robot platforms and dexterous hands for manipulation-heavy service, home, logistics and research workflows.