Closing the Lab-to-Store Gap: A Data-Efficient Post-Training and Experience-Driven Learning VLA Framework for Retail Humanoids
Closing the Lab-to-Store Gap: A Data-Efficient Post-Training and Experience-Driven Learning VLA Framework for Retail Humanoids
This research introduces DEED, a systems-level approach to improve Vision-Language-Action (VLA) humanoid robots for real-world retail tasks. Evaluated on a supermarket chip-restocking task using a Unitree G1-Edu robot and the GR00T N1.6 foundation model, the framework includes a data-efficient post-training pipeline, experience-driven refinement adapted from RECAP, and a latent-space analysis tool. Results indicate that bridging the gap between benchmark performance and reliable operation is primarily a systems integration challenge, where careful data design and targeted post-training can transform a policy failing under naive fine-tuning into a competent real-world system using only a single GPU.
The paper proposes DEED, a data-efficient post-training framework evaluated on a supermarket chip-restocking task using a Unitree G1-Edu humanoid and the GR00T N1.6 model, concluding that bridging the lab-to-store gap is primarily a systems integration challenge involving careful data design rather than an architectural one.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
on a supermarket chip-restocking task using a Unitree G1-Edu humanoid robot and the GR00T N1.6 foundation model. DEED comprises three key components: (1) a data-efficient post-training pipeline with control-frequency alignment, data curation, task-relevant visual highlighting, and reduced VLA dependence; (2) a real-world study of experience-driven refinement, adapted from RECAP via a text-based advantage prefix and
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org