A Unified Benchmark for Dynamic Medical Treatment Reinforcement Learning
Paper • 2606.01028 • 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 84, 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.
MedGym
This repository contains datasets, checkpoints, and other large artifacts used in the MedGym benchmark.
For documentation, benchmark setup, evaluation protocols, and code, please refer to the main GitHub repository:
https://github.com/kawanoKen/Medgym
Contents
| Archive | Size | Description |
|---|---|---|
checkpoints-cohort.tar.zst |
~1.4 GB | PINN and online policy checkpoints for 7 sepsis cohorts (cohort_1 – cohort_7); cohort_7 also includes pre-trained offline policies |
checkpoints-remaining15.tar.zst |
~300 MB | PINN and online/offline policy checkpoints for the 15-patient medoid/nearest-patient clustering experiments |
checkpoints-hypotension.tar.zst |
~42 MB | PINN and policy checkpoints for the acute hypotension benchmark (ind / clu / pop, fixdt and vardt) |
checkpoints-case-pinn.tar.zst |
~4 MB | PINN simulators and pre-computed rollouts for the patient 202347 case study |
Directory layouts for each archive are documented in the GitHub repository README.
Download
hf download Kenkawano/medgym \
--repo-type dataset \
--local-dir .
for f in checkpoints-*.tar.zst; do tar --zstd -xf "$f"; done
Extract the archives at the root of the GitHub repository so that the evaluation scripts find them at their default paths.
Data Availability
These checkpoints were derived from MIMIC-III and the PhysioNet Health Gym — Synthetic Acute Hypotension dataset. Neither source dataset is redistributed here; running the released evaluations does not require them. See the GitHub repository for preprocessing details.
Citation
If you use MedGym in your research, please cite:
@misc{wang2026medgym,
title = {{MedGym}: A Unified Continuous-Time Benchmark for Dynamic Medical Treatment Reinforcement Learning},
author = {Wang, Yuepeng and Kawano, Ken and Fujisawa, Yoshihiko and Zhou, Yongqi and Wachi, Akifumi and Sadria, Mehrshad and Zhou, Lei and Weiss, Richard and Fujisawa, Katsuki and Chen, Ying and Liu, Xin and Kim, Kyoung-Sook and Hu, Xiao and Gros, Sebastien and Shen, Xun},
year = {2026},
eprint = {2606.01028},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
doi = {10.48550/arXiv.2606.01028},
url = {https://arxiv.org/abs/2606.01028}
}
- Downloads last month
- 102