PhysiCalWorld: In-Context Physics Calibration for World Models
Paper • 2610.30125 • Published
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
PhysiCalWorld Datasets
Simulation datasets for PhysiCalWorld: In-Context Physics Calibration for World Models. In each environment, physical parameters that cannot be inferred from a single frame vary across 150 configurations: 135 for training and 15 held out for evaluation.
| Environment | Varied physics | Episodes | Archives |
|---|---|---|---|
| Push-T | Center of mass of the T-block | 15,150 | tar/pusht/ |
| Reacher | Damping of both joints | 3,000 | tar/reacher/ |
| BeerPong | Table contact damping | 9,000 | tar/beerpong/ |
| FetchSlide | Puck mass, table friction, and air drag | 36,000 | tar/fetch_slide/ |
Each environment provides RGB observations, actions, and states as Zarr archives, together with segmentation masks used to compute foreground IoU.
Two additional subsets are used for the paper's analyses:
tar/analysis/pusht_context/: Push-T interaction clips for 100 centers of mass, used for the descriptor similarity analysis.tar/analysis/reacher_extrapolation/: Reacher episodes whose first-joint damping lies inside and outside the training range, used for the extrapolation analysis.
Usage
Download this repository into the code repository's data/ directory and extract the archives:
pixi run hf download physical-world/physical-world-dataset-release --repo-type dataset --local-dir data
bash data/untar_from_hf.sh
Paper: https://arxiv.org/abs/2610.30125 · Code: https://github.com/physical-world/physical-world-release · Project page: https://physical-world.github.io/
The BeerPong environment uses MJCF and mesh assets from fancy_gym (MIT License).
- Downloads last month
- 31