microsoft/BusyBox

BusyBox Dataset

Description

The BusyBox dataset is a collection of language-annotated demonstration trajectories for various tasks from the BusyBox benchmark. It has flavors collected on Trossen Aloha Mobile and I2RT YAM Box robots. Data from all embodiments is in the LeRobot format and comprises several hours of teleoperated demonstrations for the following BusyBox tasks:

• —Setting sliders to specified positions
• —Pulling out and inserting wires with barrel connectors
• —Rotating a knob to a specified position
• —Flipping switches on and off
• —Reorienting and moving the BusyBox
• —Pressing buttons of various colors

Sample demonstrations of these tasks are shown in the following video:

<video src="https://microsoft.github.io/BusyBox/BusyBoxassets/bbteleop.mp4" width="640" height="480" controls></video>

BusyBox project page: https://github.com/microsoft/BusyBox

Paper: see the project webpage

Uses

This dataset can be used for finetuning VLAs for the BusyBox tasks as well as in earlier VLA training stages for enhancing the VLA's general capabilities at controlling bimanual robots.

Contacts

Galen Mullins, Andrey Kolobov

Citation

@misc{busybox2025,
  title={Benchmarking Affordance Generalization with BusyBox},
  author={Fortier, Dean and Adamson, Timothy and Hellebrekers, Tess and LaScala, Teresa and Ennin, Kofi and Murray, Michael and Kolobov, Andrey and Mullins, Galen},
  booktitle={Eval&Deploy Workshop at CoRL-2025: Evaluation and Deployment Across the Robot Learning Lifecycle},
  year={2025}
}