The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: TypeError
Message: Couldn't cast array of type
struct<A: string, B: string, C: string, D: string, E: string, F: string, G: string, H: string, I: string, J: string>
to
{'A': Value('string'), 'B': Value('string'), 'C': Value('string'), 'D': Value('string'), 'E': Value('string')}
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2312, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2158, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
struct<A: string, B: string, C: string, D: string, E: string, F: string, G: string, H: string, I: string, J: string>
to
{'A': Value('string'), 'B': Value('string'), 'C': Value('string'), 'D': Value('string'), 'E': Value('string')}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.
Public Agentic Medical Benchmarks
Download and restore large payloads
Large payloads are stored under archive-parts/ as 4 MiB LFS blocks to avoid unreliable large multipart transfers. PAYLOADS.json records every block hash and the exact original payload hash. Download this dataset repository with LFS payloads, then run:
python3 assemble_payloads.py
The script verifies each block and verifies that the reassembled archive or data file is byte-for-byte identical to the reviewed original. You can restore just one payload with --payload artifacts/FILE.tar.gz. Run the existing restoration commands below after reassembly. Original payload paths documented below refer to these reassembled files.
A consolidated collection of medical evaluation datasets, runner code, and saved benchmark evidence. This is an evaluation archive, not a training corpus or a clinical service. Preserve all upstream benchmark contamination markers.
Datasets
| Benchmark | Scope | Upstream |
|---|---|---|
| MedXpertQA-MM | 2,000 test + 5 dev; accompanying images | TsinghuaC3I/MedXpertQA |
| MedXpertQA-Text | 2,450 test + 5 dev | TsinghuaC3I/MedXpertQA |
| HealthBench Hard | 1,000 conversations and rubrics | openai/healthbench |
| HealthBench Professional | 525 conversations and rubrics | openai/healthbench-professional |
| AgentClinic MedQA-Ext | 214 simulated cases | SamuelSchmidgall/AgentClinic |
The dataset directories preserve upstream source revisions and attribution. Each component retains its own license; no blanket license replaces upstream terms. These five datasets are described as MIT-licensed by their upstream cards/repository; the bundled source notices remain authoritative. dataset-manifest.json records the exact released file hashes.
Full-scope results table · Machine-readable results index
Evaluation interpretation
Saved runs may include retries, reused answers, custom judges, failed requests, or mixed deployments. A completed campaign is not necessarily an error-free run. HealthBench rubric scores are not multiple-choice accuracy. Read the run metadata before comparing configurations. Evaluation questions, reference answers, rubrics and trajectories must remain outside training data.
Privacy and reproduction
The public export excludes credentials, private connection settings, raw client session homes, installed environments and machine caches. Local identifiers and infrastructure details in derived artifacts are sanitized. Configure your own model endpoint and credentials; historical service access is not included. Model-generated medical answers are evaluation artifacts and may be incorrect.
The source manifests and export manifests identify the exact snapshots and released artifact hashes.
Restore selected artifacts
Run python restore.py --target ./restored --only BFCL (cowork) or omit --only to restore all available archives and medical datasets. Every extracted member is checked against its hash manifest. Medical image copies can be deduplicated with --images. The script reconstructs shared runner links. Historical machine paths are marked {USER_HOME}; configure local paths and your own provider settings before executing saved runners.
- Downloads last month
- 251