Hi! I'm Niels from the community science team at Hugging Face. I've noticed this dataset repository is currently empty, so I've opened this PR to add a dataset card based on the information provided in the paper "Scaling Short-Term Memory of Visuomotor Policies for Long-Horizon Tasks" and the associated GitHub repository.
This PR adds:
- Metadata for the `robotics` task category and relevant tags.
- Links to the paper, project page, and code.
- A detailed overview of the eight task categories in the benchmark.
- Instructions for downloading the dataset via the Hugging Face CLI.
- Documentation of the HDF5 file structure (observations, actions, images).
- Sample usage for replaying and visualizing the demonstrations.
